brain_anatomy_schizophrenia_kenza-ily.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "<div style=\"text-align: left;\">\n",
    "<table style=\"width:100%; background-color:transparent;\">\n",
    "  <tr style=\"background-color:transparent;\">\n",
    "    <td align=\"center\"; style=\"background-color:transparent;\"><a href = \"https://joliot.cea.fr/drf/joliot/recherche/neurospin\"><img src=\"https://baobablab.github.io/bhb/images/collaborators/cea.jpg\" width=\"35%\"></td>\n",
    "    <td align=\"center\"; style=\"background-color:transparent; width: 50%;\"><a href = \"https://dataia.eu/\"><img src=\"https://github.com/ramp-kits/brain_anatomy_schizophrenia/raw/main/DATAIA-h.png\" width=\"80%\"></a></td>\n",
    "  </tr>\n",
    "</table> \n",
    "</div>\n",
    "\n",
    "<center><h1>Predict schizophrenia using brain anatomy</h1></center>\n",
    "\n",
    "<center><i>Edouard Duchesnay, Antoine Grigis, Benoît Dufumier (Université Paris-Saclay, CEA, NeuroSpin), François Caud, Alexandre Gramfort (Université Paris-Saclay, DATAIA) </i></center>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pip install xgboost\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Standard library imports\n",
    "import os\n",
    "import time\n",
    "\n",
    "# Third-party library imports for data manipulation and mathematical operations\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "# Plotting libraries and related imports\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.colors\n",
    "from cycler import cycler\n",
    "import seaborn as sns\n",
    "\n",
    "# Statsmodels for statistical modeling\n",
    "import statsmodels.api as sm\n",
    "import statsmodels.formula.api as smf\n",
    "import statsmodels.stats.api as sms\n",
    "\n",
    "# Scikit-learn imports for preprocessing, model selection, and model evaluation\n",
    "from sklearn.preprocessing import LabelEncoder, StandardScaler\n",
    "from sklearn.model_selection import GridSearchCV, cross_validate, cross_val_predict, StratifiedGroupKFold\n",
    "from sklearn.metrics import balanced_accuracy_score, roc_auc_score, precision_score, recall_score, f1_score\n",
    "import sklearn.metrics as metrics\n",
    "\n",
    "# Scikit-learn imports for machine learning models\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier, XGBClassifier\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.pipeline import Pipeline, make_pipeline\n",
    "from sklearn.model_selection import StratifiedKFold, StratifiedGroupKFold\n",
    "\n",
    "# Base classes for custom transformers and estimators in scikit-learn\n",
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "\n",
    "\n",
    "import problem\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "# Unused\n",
    "# from sklearn.inspection import DecisionBoundaryDisplay\n",
    "# from sklearn.svm import LinearSVC\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# I. Descriptive statistics\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### a. Participants\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Extract participants df\n",
    "participants_train = pd.read_csv(os.path.join(\"data\", \"train_participants.csv\"))\n",
    "participants_train[\"set\"] = \"train\"\n",
    "participants_test = pd.read_csv(os.path.join(\"data\", \"test_participants.csv\"))\n",
    "participants_test[\"set\"] = \"test\"\n",
    "participants = pd.concat([participants_train, participants_test])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['participant_id', 'session', 'sex', 'age', 'diagnosis', 'study', 'site',\n",
       "       'TIV', 'CSF_Vol', 'GM_Vol', 'WM_Vol', 'set'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "participants.columns\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe thead tr:last-of-type th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th colspan=\"8\" halign=\"left\">age</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>mean</th>\n",
       "      <th>std</th>\n",
       "      <th>min</th>\n",
       "      <th>25%</th>\n",
       "      <th>50%</th>\n",
       "      <th>75%</th>\n",
       "      <th>max</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>set</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>test</th>\n",
       "      <td>103.0</td>\n",
       "      <td>33.466019</td>\n",
       "      <td>12.809935</td>\n",
       "      <td>14.0</td>\n",
       "      <td>23.5</td>\n",
       "      <td>31.0</td>\n",
       "      <td>41.0</td>\n",
       "      <td>66.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>410.0</td>\n",
       "      <td>33.068293</td>\n",
       "      <td>12.562225</td>\n",
       "      <td>15.0</td>\n",
       "      <td>22.0</td>\n",
       "      <td>29.0</td>\n",
       "      <td>43.0</td>\n",
       "      <td>65.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         age                                                    \n",
       "       count       mean        std   min   25%   50%   75%   max\n",
       "set                                                             \n",
       "test   103.0  33.466019  12.809935  14.0  23.5  31.0  41.0  66.0\n",
       "train  410.0  33.068293  12.562225  15.0  22.0  29.0  43.0  65.0"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>age</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>set</th>\n",
       "      <th>sex</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">test</th>\n",
       "      <th>0.0</th>\n",
       "      <td>65</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <td>38</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th rowspan=\"2\" valign=\"top\">train</th>\n",
       "      <th>0.0</th>\n",
       "      <td>257</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1.0</th>\n",
       "      <td>153</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           age\n",
       "set   sex     \n",
       "test  0.0   65\n",
       "      1.0   38\n",
       "train 0.0  257\n",
       "      1.0  153"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot age distribution\n",
    "\n",
    "sns.violinplot(x=\"set\", y=\"age\", hue=\"sex\", data=participants, split=True)\n",
    "plt.xticks(ticks=[0, 1], labels=[\"Train\", \"Test\"])\n",
    "plt.xlabel(\"Dataset\")\n",
    "plt.ylabel(\"Age\")\n",
    "plt.show()\n",
    "display(participants[[\"age\", \"set\"]].groupby(\"set\").describe())\n",
    "display(participants[[\"set\", \"sex\", \"age\"]].groupby([\"set\", \"sex\"]).count())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Index: 513 entries, 0 to 102\n",
      "Data columns (total 12 columns):\n",
      " #   Column          Non-Null Count  Dtype  \n",
      "---  ------          --------------  -----  \n",
      " 0   participant_id  513 non-null    object \n",
      " 1   session         513 non-null    object \n",
      " 2   sex             513 non-null    float64\n",
      " 3   age             513 non-null    float64\n",
      " 4   diagnosis       513 non-null    object \n",
      " 5   study           513 non-null    object \n",
      " 6   site            513 non-null    object \n",
      " 7   TIV             513 non-null    float64\n",
      " 8   CSF_Vol         513 non-null    float64\n",
      " 9   GM_Vol          513 non-null    float64\n",
      " 10  WM_Vol          513 non-null    float64\n",
      " 11  set             513 non-null    object \n",
      "dtypes: float64(6), object(6)\n",
      "memory usage: 52.1+ KB\n"
     ]
    }
   ],
   "source": [
    "participants.info()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "object_columns = participants.select_dtypes(include=\"object\").columns.tolist()\n",
    "object_columns = [col for col in object_columns if col != \"participant_id\"]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Unique values for session : ['V1']\n",
      "Unique values for diagnosis : ['schizophrenia' 'control']\n",
      "Unique values for study : ['SCHIZCONNECT-VIP']\n",
      "Unique values for site : ['NU' 'MRN' 'WUSTL']\n",
      "Unique values for set : ['train' 'test']\n"
     ]
    }
   ],
   "source": [
    "participants.head()\n",
    "for col in object_columns:\n",
    "    print(f\"Unique values for {col} : {participants[col].unique()}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sex</th>\n",
       "      <th>age</th>\n",
       "      <th>TIV</th>\n",
       "      <th>CSF_Vol</th>\n",
       "      <th>GM_Vol</th>\n",
       "      <th>WM_Vol</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>513.000000</td>\n",
       "      <td>513.000000</td>\n",
       "      <td>513.000000</td>\n",
       "      <td>513.000000</td>\n",
       "      <td>513.000000</td>\n",
       "      <td>513.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>0.372320</td>\n",
       "      <td>33.148148</td>\n",
       "      <td>1426.607843</td>\n",
       "      <td>284.849383</td>\n",
       "      <td>625.735858</td>\n",
       "      <td>514.981154</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.483895</td>\n",
       "      <td>12.600746</td>\n",
       "      <td>149.019602</td>\n",
       "      <td>68.807070</td>\n",
       "      <td>74.945900</td>\n",
       "      <td>63.742644</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>14.000000</td>\n",
       "      <td>1071.499793</td>\n",
       "      <td>144.985388</td>\n",
       "      <td>428.563675</td>\n",
       "      <td>359.708027</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>23.000000</td>\n",
       "      <td>1327.256106</td>\n",
       "      <td>232.636949</td>\n",
       "      <td>570.983541</td>\n",
       "      <td>472.661075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>29.000000</td>\n",
       "      <td>1420.479557</td>\n",
       "      <td>275.634286</td>\n",
       "      <td>620.268996</td>\n",
       "      <td>509.191309</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>43.000000</td>\n",
       "      <td>1531.174963</td>\n",
       "      <td>325.010302</td>\n",
       "      <td>677.852937</td>\n",
       "      <td>559.486417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>66.000000</td>\n",
       "      <td>1852.574833</td>\n",
       "      <td>570.503425</td>\n",
       "      <td>889.396408</td>\n",
       "      <td>741.492388</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              sex         age          TIV     CSF_Vol      GM_Vol      WM_Vol\n",
       "count  513.000000  513.000000   513.000000  513.000000  513.000000  513.000000\n",
       "mean     0.372320   33.148148  1426.607843  284.849383  625.735858  514.981154\n",
       "std      0.483895   12.600746   149.019602   68.807070   74.945900   63.742644\n",
       "min      0.000000   14.000000  1071.499793  144.985388  428.563675  359.708027\n",
       "25%      0.000000   23.000000  1327.256106  232.636949  570.983541  472.661075\n",
       "50%      0.000000   29.000000  1420.479557  275.634286  620.268996  509.191309\n",
       "75%      1.000000   43.000000  1531.174963  325.010302  677.852937  559.486417\n",
       "max      1.000000   66.000000  1852.574833  570.503425  889.396408  741.492388"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "participants.describe()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Unsupervized analysis: Explore main sources of variabilities\n",
    "\n",
    "Load ROIs data, `train_rois.csv` provides:\n",
    "\n",
    "- Global volumes of \"tissues\": CerrebroSpinal Fluid (`CSF_Vol`), Grey (`GM_Vol`) and White Matter (`WM_Vol`) volume of participants.\n",
    "- ROIs are starting at column `l3thVen_GM_Vol`. Note that `rois_train.loc[:, 'l3thVen_GM_Vol':]` matches `problem.get_train_data()[:, :284]`.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### b. ROIS\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "rois_train = pd.read_csv(os.path.join(\"data\", \"train_rois.csv\")).loc[\n",
    "    :, \"l3thVen_GM_Vol\":\n",
    "]\n",
    "rois_test = pd.read_csv(os.path.join(\"data\", \"test_rois.csv\")).loc[:, \"l3thVen_GM_Vol\":]\n",
    "rois = pd.concat([rois_train, rois_test])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "284"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(rois_train.columns)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "PCA on ROIs: explore global effect of age\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sum_sq</th>\n",
       "      <th>df</th>\n",
       "      <th>F</th>\n",
       "      <th>PR(&gt;F)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>diagnosis</th>\n",
       "      <td>2606.207928</td>\n",
       "      <td>1.0</td>\n",
       "      <td>29.245287</td>\n",
       "      <td>9.808873e-08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Residual</th>\n",
       "      <td>45538.013350</td>\n",
       "      <td>511.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 sum_sq     df          F        PR(>F)\n",
       "diagnosis   2606.207928    1.0  29.245287  9.808873e-08\n",
       "Residual   45538.013350  511.0        NaN           NaN"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>sum_sq</th>\n",
       "      <th>df</th>\n",
       "      <th>F</th>\n",
       "      <th>PR(&gt;F)</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>diagnosis</th>\n",
       "      <td>791.094363</td>\n",
       "      <td>1.0</td>\n",
       "      <td>10.876419</td>\n",
       "      <td>0.001042</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Residual</th>\n",
       "      <td>37167.493121</td>\n",
       "      <td>511.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 sum_sq     df          F    PR(>F)\n",
       "diagnosis    791.094363    1.0  10.876419  0.001042\n",
       "Residual   37167.493121  511.0        NaN       NaN"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# PCA: Principal Component Analysis\n",
    "PCs = PCA(n_components=2).fit_transform(rois)  # Fitting on ROIs\n",
    "participants[\"PC1\"], participants[\"PC2\"] = (\n",
    "    PCs[:, 0],\n",
    "    PCs[:, 1],\n",
    ")  # Adding the PCs to the participants df\n",
    "\n",
    "sns.scatterplot(x=\"PC1\", y=\"PC2\", hue=\"diagnosis\", data=participants)\n",
    "\n",
    "\n",
    "# smf.ols\n",
    "oneway = smf.ols(\"PC1 ~ diagnosis\", participants).fit()  # Define the fitted model\n",
    "display(\n",
    "    sm.stats.anova_lm(oneway, typ=2)\n",
    ")  # ANOVA test on a lineal model - type 2:tests whether the mean of a dependent variable is different\n",
    "# across  different levels of a categorical variable.\n",
    "# smf.ols -> Describe the relationship between the PC1 and PC2, and the diagnosis, i.e. predicting PC1/PC2 as a function of diagnosis\n",
    "oneway = smf.ols(\"PC2 ~ diagnosis\", participants).fit()\n",
    "display(sm.stats.anova_lm(oneway, typ=2))\n",
    "\n",
    "# SMF: StatsModels Formula API\n",
    "# OLS: Ordinary Least Squares\n",
    "# ANOVA: analysis of Variance test on the fitted model\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The columns displayed in the output of the ANOVA test are:\n",
    "\n",
    "1. `sum_sq`: This is the sum of squares. It represents the total variation in the dependent variable.\n",
    "\n",
    "2. `df`: This stands for degrees of freedom. It's the number of values in the final calculation of a statistic that are free to vary.\n",
    "\n",
    "3. `F`: This is the F statistic, a value you get when you run an ANOVA test or a regression analysis to find out if the means between two populations are significantly different.\n",
    "\n",
    "4. `PR(>F)`: This is the p-value associated with the above F statistic. It tells you the probability of your null hypothesis being true. If this value is less than your significance level (commonly 0.05), you would reject the null hypothesis and conclude that the means are significantly different.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Diagnosis is associated with two first components, we expect to be abble to predict the diagnosis\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# II. Machine Learning\n",
    "\n",
    "Import and read data\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load concatenated data: 284 features from ROIs and 331,695 features from VBM.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train = problem.get_train_data()\n",
    "X_test, y_test = problem.get_test_data()\n",
    "\n",
    "assert X_train.shape[1] == 284 + 331695\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature extractor of ROIs or voxels within the brain (VBM)\n",
    "\n",
    "Selecting only rois or vbm images:\n",
    "\n",
    "This can be achieved by a `ROIsFeatureExtractor` or `VBMFeatureExtractor`\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(410, 284)\n",
      "(410, 331695)\n"
     ]
    }
   ],
   "source": [
    "class ROIsFeatureExtractor(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"Select only the 284 ROIs features:\"\"\"\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        return self\n",
    "\n",
    "    def transform(self, X):\n",
    "        return X[:, :284]\n",
    "\n",
    "\n",
    "class VBMFeatureExtractor(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"Select only the 284 ROIs features:\"\"\"\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        return self\n",
    "\n",
    "    def transform(self, X):\n",
    "        return X[:, 284:]\n",
    "\n",
    "\n",
    "fe = ROIsFeatureExtractor()\n",
    "print(fe.transform(X_train).shape)\n",
    "\n",
    "fe = VBMFeatureExtractor()\n",
    "print(fe.transform(X_train).shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pipeline 1: Baseline\n",
    "\n",
    "**Design of predictors and their evaluation using CV and test set**\n",
    "\n",
    "The framework is evaluated with a cross-validation approach. The metrics used are the ROC-AUC (auc) and the balanced Accuracy (bacc).\n",
    "\n",
    "First we propose a simple classification predictor based on ROIs features only:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "participants_train[\"site\"].nunique()\n",
    "# Create an instance of LabelEncoder\n",
    "encoder = LabelEncoder()\n",
    "\n",
    "# Fit and transform the 'site' column in participants_train\n",
    "participants_train[\"site_encoded\"] = encoder.fit_transform(participants_train[\"site\"])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "groups = np.array(participants_train[\"site_encoded\"].astype(\"int\"))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Feature Engineering\n",
    "\n",
    "estimator = make_pipeline(ROIsFeatureExtractor(), StandardScaler())\n",
    "\n",
    "# Encode the labels to integers\n",
    "label_encoder = LabelEncoder()\n",
    "y_train_encoded = label_encoder.fit_transform(y_train)\n",
    "y_test_encoded = label_encoder.transform(y_test)\n",
    "\n",
    "# Fit the feature extractor to the training data and transform the training data\n",
    "X_train_roi = estimator.fit_transform(X_train, y_train_encoded)\n",
    "X_test_roi = estimator.transform(X_test)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Function to perform model evaluation for a single model\n",
    "def evaluate_model(\n",
    "    model,\n",
    "    f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=False,\n",
    "):\n",
    "    \"\"\"\n",
    "    Evaluates a single model with a specific feature extractor and hyperparameters, providing scores before and after fitting.\n",
    "\n",
    "    Args:\n",
    "    - model: The model instance to be evaluated.\n",
    "    - param_grid (dict): A dictionary of hyperparameter grids for the model.\n",
    "    - cv: Cross-validation splitting strategy.\n",
    "    - f_extractor: The feature extractor instance.\n",
    "    - X_train (DataFrame): Training feature dataset.\n",
    "    - y_train (Series): Training target dataset.\n",
    "    - X_test (DataFrame): Test feature dataset.\n",
    "    - y_test (Series): Test target dataset.\n",
    "\n",
    "    Returns:\n",
    "    - result_df (DataFrame): DataFrame containing model evaluation results.\n",
    "    \"\"\"\n",
    "    # Setting up the pipeline with the feature extractor and model\n",
    "    estimator = make_pipeline(f_extractor, model)\n",
    "\n",
    "    # stratified_cv = (\n",
    "    #     StratifiedKFold(n_splits=5, shuffle=True, random_state=42)\n",
    "    #     if groups == None\n",
    "    #     else StratifiedGroupKFold(n_splits=3).split(\n",
    "    #         X_train, y_train, groups=np.array(participants_train[\"site\"].astype(\"int\"))\n",
    "    #     )\n",
    "    # )\n",
    "    # cv = stratified_cv\n",
    "\n",
    "    # Performing cross-validation to get initial scores\n",
    "    if groups is not None:\n",
    "        cv = StratifiedGroupKFold(n_splits=3, shuffle=True).split(\n",
    "            X_train,\n",
    "            y_train,\n",
    "            groups=np.array(participants_train[\"site_encoded\"].astype(\"int\")),\n",
    "        )\n",
    "\n",
    "        cv_results = cross_validate(\n",
    "            estimator,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "            groups=groups,\n",
    "        )\n",
    "    else:\n",
    "        cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)\n",
    "        cv_results = cross_validate(\n",
    "            estimator,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "        )\n",
    "    print(\"# 5CV\")\n",
    "    print(\n",
    "        \"bACC=%.2f\" % cv_results[\"test_balanced_accuracy\"].mean(),\n",
    "        \"ROC-AUC=%.2f\" % cv_results[\"test_roc_auc\"].mean(),\n",
    "    )\n",
    "\n",
    "    # Refit on all train\n",
    "    estimator.fit(X_train, y_train)\n",
    "    # Apply on test\n",
    "    y_pred_train = estimator.predict(X_train)\n",
    "    y_pred_test = estimator.predict(X_test)\n",
    "    score_pred_test = estimator.predict_proba(X_test)[:, 1]\n",
    "\n",
    "    bacc_test = metrics.balanced_accuracy_score(y_test, y_pred_test)\n",
    "    auc_test = metrics.roc_auc_score(y_test, score_pred_test)\n",
    "    print(\"# Test\")\n",
    "    print(\"bACC=%.2f\" % bacc_test, \"ROC-AUC=%.2f\" % auc_test)\n",
    "\n",
    "    # Creating a DataFrame from the results\n",
    "    result_df = pd.DataFrame(\n",
    "        {\n",
    "            \"CV Balanced Accuracy\": cv_results[\"test_balanced_accuracy\"].mean(),\n",
    "            \"CV ROC-AUC\": cv_results[\"test_roc_auc\"].mean(),\n",
    "            \"Test Balanced Accuracy\": bacc_test,\n",
    "            \"Test ROC-AUC\": auc_test,\n",
    "        },\n",
    "        index=[0],\n",
    "    )\n",
    "\n",
    "    return result_df\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  1. Linear Methods\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### a. Logistic Regression\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- Logistic Regression / Definition\n",
    "model_linear = LogisticRegression(\n",
    "    solver=\"saga\",\n",
    "    max_iter=10000,\n",
    ")\n",
    "\n",
    "\n",
    "param_grid_linear = {\n",
    "    \"logisticregression__C\": [0.1, 1, 10],\n",
    "    \"logisticregression__penalty\": [\"l1\", \"l2\"],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.68 ROC-AUC=0.73\n",
      "# Test\n",
      "bACC=0.72 ROC-AUC=0.82\n",
      "Elapsed time: 16.385972023010254\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / ROI / StratKFold\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_roi = evaluate_model(\n",
    "    model_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.68312</td>\n",
       "      <td>0.731026</td>\n",
       "      <td>0.720265</td>\n",
       "      <td>0.823864</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0               0.68312    0.731026                0.720265      0.823864"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_roi\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.67 ROC-AUC=0.73\n",
      "# Test\n",
      "bACC=0.72 ROC-AUC=0.82\n",
      "Elapsed time: 12.561697006225586\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / ROI / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_roi_group_site = evaluate_model(\n",
    "    model=model_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.665399</td>\n",
       "      <td>0.727114</td>\n",
       "      <td>0.720265</td>\n",
       "      <td>0.823864</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.665399    0.727114                0.720265      0.823864"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_roi_group_site\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.71 ROC-AUC=0.80\n",
      "# Test\n",
      "bACC=0.70 ROC-AUC=0.79\n",
      "Elapsed time for Voting Classifier: 5382.334787130356\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / VBM / StratKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_vbm = evaluate_model(\n",
    "    model_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.714348</td>\n",
       "      <td>0.797125</td>\n",
       "      <td>0.698106</td>\n",
       "      <td>0.790152</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.714348    0.797125                0.698106      0.790152"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_vbm\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.68 ROC-AUC=0.77\n",
      "# Test\n",
      "bACC=0.70 ROC-AUC=0.79\n",
      "Elapsed time: 3021.6112191677094\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / VBM / StratGroupKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_vbm_group = evaluate_model(\n",
    "    model_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.679833</td>\n",
       "      <td>0.769927</td>\n",
       "      <td>0.698106</td>\n",
       "      <td>0.79053</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.679833    0.769927                0.698106       0.79053"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_vbm_group\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### b. LinearSVC\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- Linear SVC / Definition\n",
    "model_linearsvc = SVC(kernel=\"linear\", probability=True)  # Necessary for 'l1' penalty\n",
    "\n",
    "\n",
    "param_grid_linear_svc = {\n",
    "    \"linearsvc__C\": [0.1, 1, 10],\n",
    "    \"linearsvc__penalty\": [\"l1\", \"l2\"],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.63 ROC-AUC=0.69\n",
      "# Test\n",
      "bACC=0.71 ROC-AUC=0.80\n",
      "Elapsed time: 6.450428009033203\n"
     ]
    }
   ],
   "source": [
    "# ---------- Linear SVC / ROI / StratKFold\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linearsvc_roi = evaluate_model(\n",
    "    model_linearsvc,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.632036</td>\n",
       "      <td>0.685022</td>\n",
       "      <td>0.713826</td>\n",
       "      <td>0.801894</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.632036    0.685022                0.713826      0.801894"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linearsvc_roi\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.60 ROC-AUC=0.68\n",
      "# Test\n",
      "bACC=0.71 ROC-AUC=0.80\n",
      "Elapsed time: 3.7978150844573975\n"
     ]
    }
   ],
   "source": [
    "# ---------- Linear SVC / ROI / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linearsvc_roi_group_site = evaluate_model(\n",
    "    model_linearsvc,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.597681</td>\n",
       "      <td>0.675721</td>\n",
       "      <td>0.713826</td>\n",
       "      <td>0.801894</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.597681    0.675721                0.713826      0.801894"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linearsvc_roi_group_site\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "No VBM with SVC Linear beacause of hugigh dimens\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Tree-based\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####  a. RandomForest\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- RandomForest / Definitions\n",
    "\n",
    "# RandomForest model\n",
    "model_rf = RandomForestClassifier()\n",
    "\n",
    "param_grid_rf = {\n",
    "    \"randomforestclassifier__n_estimators\": [100, 200, 300],\n",
    "    \"randomforestclassifier__max_depth\": [None, 5, 10],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.73 ROC-AUC=0.78\n",
      "# Test\n",
      "bACC=0.73 ROC-AUC=0.80\n",
      "Elapsed time for Voting Classifier: 8.55095911026001\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / ROI / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_roi = evaluate_model(\n",
    "    model_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time for Voting Classifier: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.73178</td>\n",
       "      <td>0.776358</td>\n",
       "      <td>0.72822</td>\n",
       "      <td>0.795455</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0               0.73178    0.776358                 0.72822      0.795455"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_roi\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.64 ROC-AUC=0.72\n",
      "# Test\n",
      "bACC=0.73 ROC-AUC=0.82\n",
      "Elapsed time for Voting Classifier: 3.401211977005005\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / ROI / StratGroupKFold\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_roi_groups = evaluate_model(\n",
    "    model_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time for Voting Classifier: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.644894</td>\n",
       "      <td>0.722354</td>\n",
       "      <td>0.734659</td>\n",
       "      <td>0.824811</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.644894    0.722354                0.734659      0.824811"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_roi_groups\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.69 ROC-AUC=0.76\n",
      "# Test\n",
      "bACC=0.63 ROC-AUC=0.69\n",
      "Elapsed time: 33.733633041381836\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / VBM / StratKFold\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_vbm = evaluate_model(\n",
    "    model_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n",
    "\n",
    "# ! Update here too in the report\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.694163</td>\n",
       "      <td>0.757559</td>\n",
       "      <td>0.631818</td>\n",
       "      <td>0.689962</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.694163    0.757559                0.631818      0.689962"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_vbm\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.64 ROC-AUC=0.74\n",
      "# Test\n",
      "bACC=0.60 ROC-AUC=0.70\n",
      "Elapsed time: 22.779906749725342\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / VBM / StratGroupKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_vbm_groups = evaluate_model(\n",
    "    model_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.642184</td>\n",
       "      <td>0.738983</td>\n",
       "      <td>0.60322</td>\n",
       "      <td>0.698485</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.642184    0.738983                 0.60322      0.698485"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_vbm_groups\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "####  b. XGBoost\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- XGBoost / Definitions\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "model_xgboost = XGBClassifier()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.74 ROC-AUC=0.82\n",
      "# Test\n",
      "bACC=0.75 ROC-AUC=0.84\n",
      "Elapsed time for Voting Classifier: 4.742335081100464\n"
     ]
    }
   ],
   "source": [
    "# ---------- XGBoost / ROI / StratKFold\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_xgb_roi = evaluate_model(\n",
    "    model_xgboost,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time for Voting Classifier: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.737285</td>\n",
       "      <td>0.821701</td>\n",
       "      <td>0.754167</td>\n",
       "      <td>0.8375</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.737285    0.821701                0.754167        0.8375"
      ]
     },
     "execution_count": 190,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_xgb_roi\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.70 ROC-AUC=0.77\n",
      "# Test\n",
      "bACC=0.75 ROC-AUC=0.84\n",
      "Elapsed time for Voting Classifier: 3.137244939804077\n"
     ]
    }
   ],
   "source": [
    "# ---------- XGBoost / ROI / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_xgb_roi_groups = evaluate_model(\n",
    "    model_xgboost,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time for Voting Classifier: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.701943</td>\n",
       "      <td>0.766923</td>\n",
       "      <td>0.754167</td>\n",
       "      <td>0.8375</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.701943    0.766923                0.754167        0.8375"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_xgb_roi_groups\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.72 ROC-AUC=0.81\n",
      "# Test\n",
      "bACC=0.70 ROC-AUC=0.80\n",
      "Elapsed time for Voting Classifier: 1615.7503590583801\n"
     ]
    }
   ],
   "source": [
    "# ---------- XGBoost / VBM / StratKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_xgb_vbm = evaluate_model(\n",
    "    model_xgboost,\n",
    "    cv=strat_cv,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.718673</td>\n",
       "      <td>0.810236</td>\n",
       "      <td>0.700758</td>\n",
       "      <td>0.795833</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.718673    0.810236                0.700758      0.795833"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_xgb_vbm\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "XGBoost with rbf kernel\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  3. Non-Linear\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### a. SVM\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- SVC, rbf kernel / Definitions\n",
    "model_svm = SVC(probability=True)\n",
    "param_grid_svm = {\n",
    "    \"C\": [0.1, 1, 10, 100],  # Example values\n",
    "    \"gamma\": [1, 0.1, 0.01, 0.001],\n",
    "    \"kernel\": [\"rbf\", \"linear\"],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.63 ROC-AUC=0.70\n",
      "# Test\n",
      "bACC=0.64 ROC-AUC=0.68\n",
      "Elapsed time: 4.510223865509033\n"
     ]
    }
   ],
   "source": [
    "# ---------- SVC, rbf kernel / ROI / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_svm_roi = evaluate_model(\n",
    "    model_svm,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style type=\"text/css\">\n",
       "#T_30809_row0_col0, #T_30809_row0_col1, #T_30809_row0_col2, #T_30809_row0_col3 {\n",
       "  background-color: #fff5f0;\n",
       "  color: #000000;\n",
       "}\n",
       "</style>\n",
       "<table id=\"T_30809\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th class=\"blank level0\" >&nbsp;</th>\n",
       "      <th id=\"T_30809_level0_col0\" class=\"col_heading level0 col0\" >CV Balanced Accuracy</th>\n",
       "      <th id=\"T_30809_level0_col1\" class=\"col_heading level0 col1\" >CV ROC-AUC</th>\n",
       "      <th id=\"T_30809_level0_col2\" class=\"col_heading level0 col2\" >Test Balanced Accuracy</th>\n",
       "      <th id=\"T_30809_level0_col3\" class=\"col_heading level0 col3\" >Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th id=\"T_30809_level0_row0\" class=\"row_heading level0 row0\" >0</th>\n",
       "      <td id=\"T_30809_row0_col0\" class=\"data row0 col0\" >0.632443</td>\n",
       "      <td id=\"T_30809_row0_col1\" class=\"data row0 col1\" >0.699470</td>\n",
       "      <td id=\"T_30809_row0_col2\" class=\"data row0 col2\" >0.643561</td>\n",
       "      <td id=\"T_30809_row0_col3\" class=\"data row0 col3\" >0.682197</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x29e67b9e0>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svm_roi.style.background_gradient(cmap=\"Reds\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.57 ROC-AUC=0.67\n",
      "# Test\n",
      "bACC=0.64 ROC-AUC=0.68\n",
      "Elapsed time: 2.299607753753662\n"
     ]
    }
   ],
   "source": [
    "# ---------- SVC, rbf kernel / ROI / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_svm_roi_groups = evaluate_model(\n",
    "    model_svm,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.570442</td>\n",
       "      <td>0.665289</td>\n",
       "      <td>0.643561</td>\n",
       "      <td>0.682197</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.570442    0.665289                0.643561      0.682197"
      ]
     },
     "execution_count": 80,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svm_roi_groups\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### b. KNN\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- KNN / Definitions\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "model_knn = KNeighborsClassifier()\n",
    "param_grid_knn = {\n",
    "    \"n_neighbors\": [3, 5, 7, 9],  # Example values for number of neighbors\n",
    "    \"weights\": [\"uniform\", \"distance\"],  # Weight options\n",
    "    \"algorithm\": [\"auto\", \"ball_tree\", \"kd_tree\", \"brute\"],  # Algorithm options\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.57 ROC-AUC=0.62\n",
      "# Test\n",
      "bACC=0.63 ROC-AUC=0.68\n",
      "Elapsed time: 6.640761852264404\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / ROI / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_roi = evaluate_model(\n",
    "    model_knn,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style type=\"text/css\">\n",
       "#T_f7eb8_row0_col0, #T_f7eb8_row0_col1, #T_f7eb8_row0_col2, #T_f7eb8_row0_col3 {\n",
       "  background-color: #fff5f0;\n",
       "  color: #000000;\n",
       "}\n",
       "</style>\n",
       "<table id=\"T_f7eb8\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th class=\"blank level0\" >&nbsp;</th>\n",
       "      <th id=\"T_f7eb8_level0_col0\" class=\"col_heading level0 col0\" >CV Balanced Accuracy</th>\n",
       "      <th id=\"T_f7eb8_level0_col1\" class=\"col_heading level0 col1\" >CV ROC-AUC</th>\n",
       "      <th id=\"T_f7eb8_level0_col2\" class=\"col_heading level0 col2\" >Test Balanced Accuracy</th>\n",
       "      <th id=\"T_f7eb8_level0_col3\" class=\"col_heading level0 col3\" >Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th id=\"T_f7eb8_level0_row0\" class=\"row_heading level0 row0\" >0</th>\n",
       "      <td id=\"T_f7eb8_row0_col0\" class=\"data row0 col0\" >0.572211</td>\n",
       "      <td id=\"T_f7eb8_row0_col1\" class=\"data row0 col1\" >0.616537</td>\n",
       "      <td id=\"T_f7eb8_row0_col2\" class=\"data row0 col2\" >0.626705</td>\n",
       "      <td id=\"T_f7eb8_row0_col3\" class=\"data row0 col3\" >0.681439</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x138072420>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_roi.style.background_gradient(cmap=\"Reds\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.52 ROC-AUC=0.52\n",
      "# Test\n",
      "bACC=0.63 ROC-AUC=0.68\n",
      "Elapsed time: 1.897535800933838\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / ROI / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_roi_groups = evaluate_model(\n",
    "    model_knn,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.51736</td>\n",
       "      <td>0.520491</td>\n",
       "      <td>0.626705</td>\n",
       "      <td>0.681439</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0               0.51736    0.520491                0.626705      0.681439"
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_roi_groups\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.56 ROC-AUC=0.61\n",
      "# Test\n",
      "bACC=0.63 ROC-AUC=0.62\n",
      "Elapsed time: 32.503901958465576\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / VBM / StratKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_vbm = evaluate_model(\n",
    "    model_knn,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=None,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n",
    "\n",
    "# ! Update here too in the report\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.561162</td>\n",
       "      <td>0.60636</td>\n",
       "      <td>0.629167</td>\n",
       "      <td>0.622159</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0              0.561162     0.60636                0.629167      0.622159"
      ]
     },
     "execution_count": 87,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_vbm\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# 5CV\n",
      "bACC=0.56 ROC-AUC=0.60\n",
      "# Test\n",
      "bACC=0.63 ROC-AUC=0.62\n",
      "Elapsed time: 20.882786989212036\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / VBM / StratGroupKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_vbm_groups = evaluate_model(\n",
    "    model_knn,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>CV Balanced Accuracy</th>\n",
       "      <th>CV ROC-AUC</th>\n",
       "      <th>Test Balanced Accuracy</th>\n",
       "      <th>Test ROC-AUC</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.56484</td>\n",
       "      <td>0.602996</td>\n",
       "      <td>0.629167</td>\n",
       "      <td>0.622159</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   CV Balanced Accuracy  CV ROC-AUC  Test Balanced Accuracy  Test ROC-AUC\n",
       "0               0.56484    0.602996                0.629167      0.622159"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_vbm_groups\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### c/Evaluation\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pipeline 2: GridSearch, regularisation and feature engineering\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Kernels\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "binary_cmap = matplotlib.colors.LinearSegmentedColormap.from_list(\n",
    "    \"\", [\"#332288\", \"white\", \"#AA4499\"]\n",
    ")\n",
    "plt.rcParams[\"axes.prop_cycle\"] = cycler(\n",
    "    color=[\n",
    "        \"#332288\",\n",
    "        \"#88CCEE\",\n",
    "        \"#44AA99\",\n",
    "        \"#117733\",\n",
    "        \"#999933\",\n",
    "        \"#DDCC77\",\n",
    "        \"#CC6677\",\n",
    "        \"#882255\",\n",
    "        \"#AA4499\",\n",
    "    ]\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "estimator = make_pipeline(ROIsFeatureExtractor(), StandardScaler())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# kernel names\n",
    "kernels = [\"linear\", \"poly\", \"rbf\", \"sigmoid\"]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import accuracy_score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Kernel: linear \n",
      "# 5CV\n",
      "bACC=0.66 ROC-AUC=0.72\n",
      "# Test\n",
      "bACC=0.68 ROC-AUC=0.75\n",
      "Elapsed time for linear kernel: 7.478082656860352 seconds\n",
      "-------------------\n",
      "Kernel: poly \n",
      "# 5CV\n",
      "bACC=0.62 ROC-AUC=0.78\n",
      "# Test\n",
      "bACC=0.58 ROC-AUC=0.78\n",
      "Elapsed time for poly kernel: 7.166824817657471 seconds\n",
      "-------------------\n",
      "Kernel: rbf \n",
      "# 5CV\n",
      "bACC=0.73 ROC-AUC=0.82\n",
      "# Test\n",
      "bACC=0.75 ROC-AUC=0.84\n",
      "Elapsed time for rbf kernel: 7.311755895614624 seconds\n",
      "-------------------\n",
      "Kernel: sigmoid \n",
      "# 5CV\n",
      "bACC=0.68 ROC-AUC=0.72\n",
      "# Test\n",
      "bACC=0.73 ROC-AUC=0.78\n",
      "Elapsed time for sigmoid kernel: 7.072613954544067 seconds\n",
      "-------------------\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import SVC\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.pipeline import make_pipeline\n",
    "\n",
    "# Kernel names\n",
    "kernels = [\"linear\", \"poly\", \"rbf\", \"sigmoid\"]\n",
    "\n",
    "# Dictionary to store results\n",
    "results = {}\n",
    "\n",
    "# Evaluating an SVC model with VBM feature extractor for each kernel\n",
    "for kernel in kernels:\n",
    "    print(f\"Kernel: {kernel} \")\n",
    "\n",
    "    # Initialize the SVM model with the current kernel and probability=True\n",
    "    model_svc = make_pipeline(StandardScaler(), SVC(kernel=kernel, probability=True))\n",
    "\n",
    "    # Initialize the feature extractor\n",
    "    f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "    # Start timing\n",
    "    start_time = time.time()\n",
    "\n",
    "    # Evaluate the model and store the result in the dictionary\n",
    "    results[f\"result_df_svc_roi_{kernel}\"] = evaluate_model(\n",
    "        model_svc,  # This is now an SVC model with the specified kernel and probability enabled\n",
    "        cv=strat_cv,\n",
    "        f_extractor=f_extractor,\n",
    "        X_train=X_train,\n",
    "        y_train=y_train,\n",
    "        X_test=X_test,\n",
    "        y_test=y_test,\n",
    "        groups=None,\n",
    "    )\n",
    "\n",
    "    # End timing\n",
    "    end_time = time.time()\n",
    "\n",
    "    # Print elapsed time\n",
    "    print(f\"Elapsed time for {kernel} kernel: {end_time - start_time} seconds\")\n",
    "    print(\"-------------------\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### GridSearch\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# #  all metrics\n",
    "# def evaluate_model_gridsearch(\n",
    "#     model,\n",
    "#     param_grid,\n",
    "#     cv,\n",
    "#     f_extractor,\n",
    "#     X_train,\n",
    "#     y_train,\n",
    "#     X_test,\n",
    "#     y_test,\n",
    "#     groups=None,\n",
    "# ):\n",
    "#     # Setting up the pipeline with the feature extractor and model\n",
    "#     estimator = make_pipeline(f_extractor, StandardScaler(), model)\n",
    "\n",
    "#     # Performing initial cross-validation to get baseline scores and fitting time\n",
    "#     cv_results = cross_validate(\n",
    "#         estimator,\n",
    "#         X_train,\n",
    "#         y_train,\n",
    "#         scoring=[\"balanced_accuracy\", \"roc_auc\", \"precision\", \"recall\", \"f1\"],\n",
    "#         cv=cv,\n",
    "#         return_train_score=True,\n",
    "#         return_estimator=True,\n",
    "#         groups=groups,\n",
    "#     )\n",
    "\n",
    "#     # Calculate baseline metrics and average fitting time\n",
    "#     baseline_metrics = {\n",
    "#         \"Balanced Accuracy\": cv_results[\"test_balanced_accuracy\"].mean(),\n",
    "#         \"ROC-AUC\": cv_results[\"test_roc_auc\"].mean(),\n",
    "#         \"Precision\": cv_results[\"test_precision\"].mean(),\n",
    "#         \"Recall\": cv_results[\"test_recall\"].mean(),\n",
    "#         \"F1 Score\": cv_results[\"test_f1\"].mean(),\n",
    "#     }\n",
    "#     baseline_fit_time = cv_results[\"fit_time\"].mean()\n",
    "\n",
    "#     # Setting up GridSearchCV to find the best parameters for the model\n",
    "#     grid_search = GridSearchCV(\n",
    "#         estimator,\n",
    "#         param_grid,\n",
    "#         scoring=\"balanced_accuracy\",\n",
    "#         refit=\"balanced_accuracy\",\n",
    "#         cv=cv,\n",
    "#         verbose=1,\n",
    "#         n_jobs=-1,\n",
    "#     )\n",
    "\n",
    "#     # Fitting GridSearchCV to find the best parameters and model, and measure fitting time\n",
    "#     grid_search.fit(X_train, y_train, groups=groups)\n",
    "#     best_model = grid_search.best_estimator_\n",
    "#     grid_search_fit_time = grid_search.cv_results_[\"mean_fit_time\"].mean()\n",
    "\n",
    "#     # Evaluating the best model on the test set\n",
    "#     y_pred_test = best_model.predict(X_test)\n",
    "#     score_pred_test = best_model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "#     # Calculating test metrics for the best model\n",
    "#     test_metrics = {\n",
    "#         \"Balanced Accuracy\": balanced_accuracy_score(y_test, y_pred_test),\n",
    "#         \"ROC-AUC\": roc_auc_score(y_test, score_pred_test),\n",
    "#         \"Precision\": precision_score(y_test, y_pred_test),\n",
    "#         \"Recall\": recall_score(y_test, y_pred_test),\n",
    "#         \"F1 Score\": f1_score(y_test, y_pred_test),\n",
    "#     }\n",
    "\n",
    "#     # Creating a DataFrame from the results including the differences and fitting times\n",
    "#     result_df = pd.DataFrame(\n",
    "#         {\n",
    "#             \"Metric\": [\n",
    "#                 \"Balanced Accuracy\",\n",
    "#                 \"ROC-AUC\",\n",
    "#                 \"Precision\",\n",
    "#                 \"Recall\",\n",
    "#                 \"F1 Score\",\n",
    "#                 \"Fit Time (s)\",\n",
    "#             ],\n",
    "#             \"Baseline\": list(baseline_metrics.values()) + [baseline_fit_time],\n",
    "#             \"GridSearch\": list(test_metrics.values()) + [grid_search_fit_time],\n",
    "#             \"Difference\": [\n",
    "#                 test_metrics[metric] - baseline_metrics[metric]\n",
    "#                 for metric in baseline_metrics.keys()\n",
    "#             ]\n",
    "#             + [grid_search_fit_time - baseline_fit_time],\n",
    "#         }\n",
    "#     )\n",
    "\n",
    "#     return result_df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train,\n",
    "    X_test,\n",
    "    y_test,\n",
    "    groups=None,\n",
    "):\n",
    "    # Setting up the pipeline with the feature extractor and model\n",
    "    estimator = make_pipeline(f_extractor, StandardScaler(), model)\n",
    "\n",
    "    # Performing initial cross-validation to get baseline scores and fitting time\n",
    "    if groups is not None:\n",
    "        cv = StratifiedGroupKFold(n_splits=3, shuffle=True).split(\n",
    "            X_train,\n",
    "            y_train,\n",
    "            groups=np.array(participants_train[\"site_encoded\"].astype(\"int\")),\n",
    "        )\n",
    "        groups = np.array(participants_train[\"site_encoded\"].astype(\"int\"))\n",
    "        cv_results = cross_validate(\n",
    "            estimator,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\", \"precision\", \"recall\", \"f1\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "            return_estimator=True,\n",
    "            groups=groups,\n",
    "        )\n",
    "    else:\n",
    "        cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)\n",
    "        cv_results = cross_validate(\n",
    "            estimator,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\", \"precision\", \"recall\", \"f1\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "            return_estimator=True,\n",
    "        )\n",
    "\n",
    "    # Calculate baseline metrics and average fitting time\n",
    "    baseline_metrics = {\n",
    "        \"Balanced Accuracy\": cv_results[\"test_balanced_accuracy\"].mean(),\n",
    "        \"ROC-AUC\": cv_results[\"test_roc_auc\"].mean(),\n",
    "        \"Precision\": cv_results[\"test_precision\"].mean(),\n",
    "        \"Recall\": cv_results[\"test_recall\"].mean(),\n",
    "        \"F1 Score\": cv_results[\"test_f1\"].mean(),\n",
    "    }\n",
    "    baseline_fit_time = cv_results[\"fit_time\"].mean()\n",
    "\n",
    "    # Setting up GridSearchCV to find the best parameters for the model\n",
    "    grid_search = GridSearchCV(\n",
    "        estimator,\n",
    "        param_grid,\n",
    "        scoring=\"balanced_accuracy\",\n",
    "        refit=\"balanced_accuracy\",\n",
    "        cv=cv,\n",
    "        verbose=1,\n",
    "        n_jobs=-1,\n",
    "    )\n",
    "\n",
    "    # Fitting GridSearchCV to find the best parameters and model, and measure fitting time\n",
    "    grid_search.fit(X_train, y_train, groups=groups)\n",
    "    best_model = grid_search.best_estimator_\n",
    "    grid_search_fit_time = grid_search.cv_results_[\"mean_fit_time\"].mean()\n",
    "\n",
    "    # Evaluating the best model on the test set\n",
    "    y_pred_test = best_model.predict(X_test)\n",
    "    score_pred_test = best_model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "    # Calculating test metrics for the best model\n",
    "    test_metrics = {\n",
    "        \"Balanced Accuracy\": balanced_accuracy_score(y_test, y_pred_test),\n",
    "        \"ROC-AUC\": roc_auc_score(y_test, score_pred_test),\n",
    "        \"Precision\": precision_score(y_test, y_pred_test),\n",
    "        \"Recall\": recall_score(y_test, y_pred_test),\n",
    "        \"F1 Score\": f1_score(y_test, y_pred_test),\n",
    "    }\n",
    "\n",
    "    # Creating a DataFrame from the results including the differences and fitting times\n",
    "    result_df = pd.DataFrame(\n",
    "        {\n",
    "            \"Metric\": [\n",
    "                \"Balanced Accuracy\",\n",
    "                \"ROC-AUC\",\n",
    "                \"Precision\",\n",
    "                \"Recall\",\n",
    "                \"F1 Score\",\n",
    "                \"Fit Time (s)\",\n",
    "            ],\n",
    "            \"Baseline\": list(baseline_metrics.values()) + [baseline_fit_time],\n",
    "            \"GridSearch\": list(test_metrics.values()) + [grid_search_fit_time],\n",
    "            \"Difference\": [\n",
    "                test_metrics[metric] - baseline_metrics[metric]\n",
    "                for metric in baseline_metrics.keys()\n",
    "            ]\n",
    "            + [grid_search_fit_time - baseline_fit_time],\n",
    "        }\n",
    "    )\n",
    "\n",
    "    return result_df\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "def evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train,\n",
    "    X_test,\n",
    "    y_test,\n",
    "    groups=None,\n",
    "):\n",
    "    # Setting up the pipeline with the feature extractor and model\n",
    "    estimator = make_pipeline(f_extractor, StandardScaler(), model)\n",
    "\n",
    "    # Setting up GridSearchCV to find the best parameters for the model\n",
    "    if groups is not None:\n",
    "        cv = StratifiedGroupKFold(n_splits=3, shuffle=True)\n",
    "    else:\n",
    "        cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)\n",
    "\n",
    "    grid_search = GridSearchCV(\n",
    "        estimator,\n",
    "        param_grid,\n",
    "        scoring=\"balanced_accuracy\",\n",
    "        refit=\"balanced_accuracy\",\n",
    "        cv=cv,\n",
    "        verbose=1,\n",
    "        n_jobs=-1,\n",
    "    )\n",
    "\n",
    "    # Fitting GridSearchCV to find the best parameters and model, and measure fitting time\n",
    "    grid_search.fit(X_train, y_train, groups=groups)\n",
    "    best_model = grid_search.best_estimator_\n",
    "    grid_search_fit_time = grid_search.cv_results_[\"mean_fit_time\"].mean()\n",
    "\n",
    "    # Performing cross-validation with the best model to get baseline scores\n",
    "    if groups is not None:\n",
    "        cv_results = cross_validate(\n",
    "            best_model,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\", \"precision\", \"recall\", \"f1\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "            return_estimator=True,\n",
    "            groups=groups,\n",
    "        )\n",
    "    else:\n",
    "        cv_results = cross_validate(\n",
    "            best_model,\n",
    "            X_train,\n",
    "            y_train,\n",
    "            scoring=[\"balanced_accuracy\", \"roc_auc\", \"precision\", \"recall\", \"f1\"],\n",
    "            cv=cv,\n",
    "            return_train_score=True,\n",
    "            return_estimator=True,\n",
    "        )\n",
    "\n",
    "    # Calculate baseline metrics and average fitting time\n",
    "    baseline_metrics = {\n",
    "        \"Balanced Accuracy\": cv_results[\"test_balanced_accuracy\"].mean(),\n",
    "        \"ROC-AUC\": cv_results[\"test_roc_auc\"].mean(),\n",
    "        \"Precision\": cv_results[\"test_precision\"].mean(),\n",
    "        \"Recall\": cv_results[\"test_recall\"].mean(),\n",
    "        \"F1 Score\": cv_results[\"test_f1\"].mean(),\n",
    "    }\n",
    "    baseline_fit_time = cv_results[\"fit_time\"].mean()\n",
    "\n",
    "    # Evaluating the best model on the test set\n",
    "    y_pred_test = best_model.predict(X_test)\n",
    "    score_pred_test = best_model.predict_proba(X_test)[:, 1]\n",
    "\n",
    "    # Calculating test metrics for the best model\n",
    "    test_metrics = {\n",
    "        \"Balanced Accuracy\": balanced_accuracy_score(y_test, y_pred_test),\n",
    "        \"ROC-AUC\": roc_auc_score(y_test, score_pred_test),\n",
    "        \"Precision\": precision_score(y_test, y_pred_test),\n",
    "        \"Recall\": recall_score(y_test, y_pred_test),\n",
    "        \"F1 Score\": f1_score(y_test, y_pred_test),\n",
    "    }\n",
    "\n",
    "    # Creating a DataFrame from the results including the differences and fitting times\n",
    "    result_df = pd.DataFrame(\n",
    "        {\n",
    "            \"Metric\": [\n",
    "                \"Balanced Accuracy\",\n",
    "                \"ROC-AUC\",\n",
    "                \"Precision\",\n",
    "                \"Recall\",\n",
    "                \"F1 Score\",\n",
    "                \"Fit Time (s)\",\n",
    "            ],\n",
    "            \"Baseline\": list(baseline_metrics.values()) + [baseline_fit_time],\n",
    "            \"GridSearch\": list(test_metrics.values()) + [grid_search_fit_time],\n",
    "            \"Difference\": [\n",
    "                test_metrics[metric] - baseline_metrics[metric]\n",
    "                for metric in baseline_metrics.keys()\n",
    "            ]\n",
    "            + [grid_search_fit_time - baseline_fit_time],\n",
    "        }\n",
    "    )\n",
    "\n",
    "    return result_df\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVC, linear\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- SVC, linear kernel / Definitions\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.model_selection import GridSearchCV, cross_validate\n",
    "from sklearn.metrics import balanced_accuracy_score, roc_auc_score\n",
    "\n",
    "\n",
    "model_svc = SVC(kernel=\"linear\", probability=True)\n",
    "\n",
    "# Define the parameter grid for the SVC\n",
    "param_grid_svc = {\"svc__C\": [0.1, 1, 10], \"svc__gamma\": [1, 0.1, 0.01]}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 9 candidates, totalling 45 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "python(20806) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20807) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20808) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20809) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20810) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20811) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20812) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20813) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20814) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20815) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20816) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(20817) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 38.391255140304565\n"
     ]
    }
   ],
   "source": [
    "# ---------- SVC, linear kernel / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_svclin_roi_gs = evaluate_model_gridsearch(\n",
    "    model_svc,\n",
    "    param_grid_svc,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.677407</td>\n",
       "      <td>0.731692</td>\n",
       "      <td>0.666071</td>\n",
       "      <td>0.61138</td>\n",
       "      <td>0.636032</td>\n",
       "      <td>1.451775</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.742424</td>\n",
       "      <td>0.831439</td>\n",
       "      <td>0.761905</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.711111</td>\n",
       "      <td>7.175701</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.065017</td>\n",
       "      <td>0.099747</td>\n",
       "      <td>0.095833</td>\n",
       "      <td>0.055287</td>\n",
       "      <td>0.075079</td>\n",
       "      <td>5.723926</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.677407  0.731692   0.666071   0.61138  0.636032   \n",
       "GridSearch           0.742424  0.831439   0.761905  0.666667  0.711111   \n",
       "Difference           0.065017  0.099747   0.095833  0.055287  0.075079   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.451775  \n",
       "GridSearch      7.175701  \n",
       "Difference      5.723926  "
      ]
     },
     "execution_count": 147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svclin_roi_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 9 candidates, totalling 27 fits\n",
      "Elapsed time: 19.759986877441406\n"
     ]
    }
   ],
   "source": [
    "# ---------- SVC, linear kernel / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_svclin_roi_gs_groups = evaluate_model_gridsearch(\n",
    "    model_svc,\n",
    "    param_grid_svc,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.621522</td>\n",
       "      <td>0.694077</td>\n",
       "      <td>0.606179</td>\n",
       "      <td>0.625774</td>\n",
       "      <td>0.607199</td>\n",
       "      <td>1.619671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.742424</td>\n",
       "      <td>0.831629</td>\n",
       "      <td>0.761905</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.711111</td>\n",
       "      <td>5.728943</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.120903</td>\n",
       "      <td>0.137552</td>\n",
       "      <td>0.155726</td>\n",
       "      <td>0.040893</td>\n",
       "      <td>0.103912</td>\n",
       "      <td>4.109272</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.621522  0.694077   0.606179  0.625774  0.607199   \n",
       "GridSearch           0.742424  0.831629   0.761905  0.666667  0.711111   \n",
       "Difference           0.120903  0.137552   0.155726  0.040893  0.103912   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.619671  \n",
       "GridSearch      5.728943  \n",
       "Difference      4.109272  "
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svclin_roi_gs_groups.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.677407</td>\n",
       "      <td>0.742424</td>\n",
       "      <td>0.065017</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.621522</td>\n",
       "      <td>0.742424</td>\n",
       "      <td>0.120903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.731692</td>\n",
       "      <td>0.831439</td>\n",
       "      <td>0.099747</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.694077</td>\n",
       "      <td>0.831629</td>\n",
       "      <td>0.137552</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.666071</td>\n",
       "      <td>0.761905</td>\n",
       "      <td>0.095833</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.606179</td>\n",
       "      <td>0.761905</td>\n",
       "      <td>0.155726</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.611380</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.055287</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.625774</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.040893</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.636032</td>\n",
       "      <td>0.711111</td>\n",
       "      <td>0.075079</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.607199</td>\n",
       "      <td>0.711111</td>\n",
       "      <td>0.103912</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.451775</td>\n",
       "      <td>7.175701</td>\n",
       "      <td>5.723926</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.619671</td>\n",
       "      <td>5.728943</td>\n",
       "      <td>4.109272</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.677407    0.742424    0.065017  Balanced Accuracy   \n",
       "1            ROC-AUC  0.731692    0.831439    0.099747            ROC-AUC   \n",
       "2          Precision  0.666071    0.761905    0.095833          Precision   \n",
       "3             Recall  0.611380    0.666667    0.055287             Recall   \n",
       "4           F1 Score  0.636032    0.711111    0.075079           F1 Score   \n",
       "5       Fit Time (s)  1.451775    7.175701    5.723926       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.621522    0.742424    0.120903  \n",
       "1  0.694077    0.831629    0.137552  \n",
       "2  0.606179    0.761905    0.155726  \n",
       "3  0.625774    0.666667    0.040893  \n",
       "4  0.607199    0.711111    0.103912  \n",
       "5  1.619671    5.728943    4.109272  "
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_svclin_roi_gs, result_df_svclin_roi_gs_groups], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "LogisticRegression\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 6 candidates, totalling 30 fits\n",
      "Elapsed time: 29.21945285797119\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_roi_gs = evaluate_model_gridsearch(\n",
    "    model=model_linear,\n",
    "    param_grid=param_grid_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.732241</td>\n",
       "      <td>0.814814</td>\n",
       "      <td>0.734677</td>\n",
       "      <td>0.675391</td>\n",
       "      <td>0.7018</td>\n",
       "      <td>1.412183</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.750189</td>\n",
       "      <td>0.842803</td>\n",
       "      <td>0.794872</td>\n",
       "      <td>0.645833</td>\n",
       "      <td>0.712644</td>\n",
       "      <td>7.477328</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.017948</td>\n",
       "      <td>0.027989</td>\n",
       "      <td>0.060194</td>\n",
       "      <td>-0.029558</td>\n",
       "      <td>0.010844</td>\n",
       "      <td>6.065146</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.732241  0.814814   0.734677  0.675391    0.7018   \n",
       "GridSearch           0.750189  0.842803   0.794872  0.645833  0.712644   \n",
       "Difference           0.017948  0.027989   0.060194 -0.029558  0.010844   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.412183  \n",
       "GridSearch      7.477328  \n",
       "Difference      6.065146  "
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_roi_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 6 candidates, totalling 18 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 13.102388858795166\n"
     ]
    }
   ],
   "source": [
    "# ---------- Logistic Regression / ROI / GridSearchCV / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_linear_roi_gs_groups_site = evaluate_model_gridsearch(\n",
    "    model=model_linear,\n",
    "    param_grid=param_grid_linear,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style type=\"text/css\">\n",
       "</style>\n",
       "<table id=\"T_6a6bf\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th class=\"blank level0\" >&nbsp;</th>\n",
       "      <th id=\"T_6a6bf_level0_col0\" class=\"col_heading level0 col0\" >0</th>\n",
       "      <th id=\"T_6a6bf_level0_col1\" class=\"col_heading level0 col1\" >1</th>\n",
       "      <th id=\"T_6a6bf_level0_col2\" class=\"col_heading level0 col2\" >2</th>\n",
       "      <th id=\"T_6a6bf_level0_col3\" class=\"col_heading level0 col3\" >3</th>\n",
       "      <th id=\"T_6a6bf_level0_col4\" class=\"col_heading level0 col4\" >4</th>\n",
       "      <th id=\"T_6a6bf_level0_col5\" class=\"col_heading level0 col5\" >5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th id=\"T_6a6bf_level0_row0\" class=\"row_heading level0 row0\" >Metric</th>\n",
       "      <td id=\"T_6a6bf_row0_col0\" class=\"data row0 col0\" >Balanced Accuracy</td>\n",
       "      <td id=\"T_6a6bf_row0_col1\" class=\"data row0 col1\" >ROC-AUC</td>\n",
       "      <td id=\"T_6a6bf_row0_col2\" class=\"data row0 col2\" >Precision</td>\n",
       "      <td id=\"T_6a6bf_row0_col3\" class=\"data row0 col3\" >Recall</td>\n",
       "      <td id=\"T_6a6bf_row0_col4\" class=\"data row0 col4\" >F1 Score</td>\n",
       "      <td id=\"T_6a6bf_row0_col5\" class=\"data row0 col5\" >Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th id=\"T_6a6bf_level0_row1\" class=\"row_heading level0 row1\" >Baseline</th>\n",
       "      <td id=\"T_6a6bf_row1_col0\" class=\"data row1 col0\" >0.711314</td>\n",
       "      <td id=\"T_6a6bf_row1_col1\" class=\"data row1 col1\" >0.800618</td>\n",
       "      <td id=\"T_6a6bf_row1_col2\" class=\"data row1 col2\" >0.722875</td>\n",
       "      <td id=\"T_6a6bf_row1_col3\" class=\"data row1 col3\" >0.662105</td>\n",
       "      <td id=\"T_6a6bf_row1_col4\" class=\"data row1 col4\" >0.683882</td>\n",
       "      <td id=\"T_6a6bf_row1_col5\" class=\"data row1 col5\" >0.821532</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th id=\"T_6a6bf_level0_row2\" class=\"row_heading level0 row2\" >GridSearch</th>\n",
       "      <td id=\"T_6a6bf_row2_col0\" class=\"data row2 col0\" >0.750189</td>\n",
       "      <td id=\"T_6a6bf_row2_col1\" class=\"data row2 col1\" >0.842803</td>\n",
       "      <td id=\"T_6a6bf_row2_col2\" class=\"data row2 col2\" >0.794872</td>\n",
       "      <td id=\"T_6a6bf_row2_col3\" class=\"data row2 col3\" >0.645833</td>\n",
       "      <td id=\"T_6a6bf_row2_col4\" class=\"data row2 col4\" >0.712644</td>\n",
       "      <td id=\"T_6a6bf_row2_col5\" class=\"data row2 col5\" >5.406044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th id=\"T_6a6bf_level0_row3\" class=\"row_heading level0 row3\" >Difference</th>\n",
       "      <td id=\"T_6a6bf_row3_col0\" class=\"data row3 col0\" >0.038875</td>\n",
       "      <td id=\"T_6a6bf_row3_col1\" class=\"data row3 col1\" >0.042185</td>\n",
       "      <td id=\"T_6a6bf_row3_col2\" class=\"data row3 col2\" >0.071997</td>\n",
       "      <td id=\"T_6a6bf_row3_col3\" class=\"data row3 col3\" >-0.016272</td>\n",
       "      <td id=\"T_6a6bf_row3_col4\" class=\"data row3 col4\" >0.028762</td>\n",
       "      <td id=\"T_6a6bf_row3_col5\" class=\"data row3 col5\" >4.584511</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x2b9353320>"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_linear_roi_gs_groups_site.T.style.background_gradient(cmap=\"Reds\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.732241</td>\n",
       "      <td>0.750189</td>\n",
       "      <td>0.017948</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.711314</td>\n",
       "      <td>0.750189</td>\n",
       "      <td>0.038875</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.814814</td>\n",
       "      <td>0.842803</td>\n",
       "      <td>0.027989</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.800618</td>\n",
       "      <td>0.842803</td>\n",
       "      <td>0.042185</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.734677</td>\n",
       "      <td>0.794872</td>\n",
       "      <td>0.060194</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.722875</td>\n",
       "      <td>0.794872</td>\n",
       "      <td>0.071997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.675391</td>\n",
       "      <td>0.645833</td>\n",
       "      <td>-0.029558</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.662105</td>\n",
       "      <td>0.645833</td>\n",
       "      <td>-0.016272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.701800</td>\n",
       "      <td>0.712644</td>\n",
       "      <td>0.010844</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.683882</td>\n",
       "      <td>0.712644</td>\n",
       "      <td>0.028762</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.412183</td>\n",
       "      <td>7.477328</td>\n",
       "      <td>6.065146</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>0.821532</td>\n",
       "      <td>5.406044</td>\n",
       "      <td>4.584511</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.732241    0.750189    0.017948  Balanced Accuracy   \n",
       "1            ROC-AUC  0.814814    0.842803    0.027989            ROC-AUC   \n",
       "2          Precision  0.734677    0.794872    0.060194          Precision   \n",
       "3             Recall  0.675391    0.645833   -0.029558             Recall   \n",
       "4           F1 Score  0.701800    0.712644    0.010844           F1 Score   \n",
       "5       Fit Time (s)  1.412183    7.477328    6.065146       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.711314    0.750189    0.038875  \n",
       "1  0.800618    0.842803    0.042185  \n",
       "2  0.722875    0.794872    0.071997  \n",
       "3  0.662105    0.645833   -0.016272  \n",
       "4  0.683882    0.712644    0.028762  \n",
       "5  0.821532    5.406044    4.584511  "
      ]
     },
     "execution_count": 113,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat(\n",
    "    [result_df_linear_roi_gs, result_df_linear_roi_gs_groups_site], axis=1\n",
    ")\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "RandomForest\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 9 candidates, totalling 45 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 30.147907972335815\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_roi_gs = evaluate_model_gridsearch(\n",
    "    model=model_rf,\n",
    "    param_grid=param_grid_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.714193</td>\n",
       "      <td>0.785575</td>\n",
       "      <td>0.719225</td>\n",
       "      <td>0.648791</td>\n",
       "      <td>0.677118</td>\n",
       "      <td>1.783538</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.737311</td>\n",
       "      <td>0.814773</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.721649</td>\n",
       "      <td>4.905923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.023117</td>\n",
       "      <td>0.029198</td>\n",
       "      <td>-0.00494</td>\n",
       "      <td>0.080376</td>\n",
       "      <td>0.044532</td>\n",
       "      <td>3.122385</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.714193  0.785575   0.719225  0.648791  0.677118   \n",
       "GridSearch           0.737311  0.814773   0.714286  0.729167  0.721649   \n",
       "Difference           0.023117  0.029198   -0.00494  0.080376  0.044532   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.783538  \n",
       "GridSearch      4.905923  \n",
       "Difference      3.122385  "
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_roi_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 9 candidates, totalling 27 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 20.585004091262817\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / ROI / GridSearchCV / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_roi_gs_groups = evaluate_model_gridsearch(\n",
    "    model=model_rf,\n",
    "    param_grid=param_grid_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.669984</td>\n",
       "      <td>0.739126</td>\n",
       "      <td>0.679976</td>\n",
       "      <td>0.64288</td>\n",
       "      <td>0.650457</td>\n",
       "      <td>1.660865</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.72822</td>\n",
       "      <td>0.792803</td>\n",
       "      <td>0.7</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>6.081828</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.058236</td>\n",
       "      <td>0.053677</td>\n",
       "      <td>0.020024</td>\n",
       "      <td>0.086286</td>\n",
       "      <td>0.063829</td>\n",
       "      <td>4.420963</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.669984  0.739126   0.679976   0.64288  0.650457   \n",
       "GridSearch            0.72822  0.792803        0.7  0.729167  0.714286   \n",
       "Difference           0.058236  0.053677   0.020024  0.086286  0.063829   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.660865  \n",
       "GridSearch      6.081828  \n",
       "Difference      4.420963  "
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_roi_gs_groups.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.714193</td>\n",
       "      <td>0.737311</td>\n",
       "      <td>0.023117</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.669984</td>\n",
       "      <td>0.728220</td>\n",
       "      <td>0.058236</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.785575</td>\n",
       "      <td>0.814773</td>\n",
       "      <td>0.029198</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.739126</td>\n",
       "      <td>0.792803</td>\n",
       "      <td>0.053677</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.719225</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>-0.004940</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.679976</td>\n",
       "      <td>0.700000</td>\n",
       "      <td>0.020024</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.648791</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.080376</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.642880</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.086286</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.677118</td>\n",
       "      <td>0.721649</td>\n",
       "      <td>0.044532</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.650457</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>0.063829</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.783538</td>\n",
       "      <td>4.905923</td>\n",
       "      <td>3.122385</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.660865</td>\n",
       "      <td>6.081828</td>\n",
       "      <td>4.420963</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.714193    0.737311    0.023117  Balanced Accuracy   \n",
       "1            ROC-AUC  0.785575    0.814773    0.029198            ROC-AUC   \n",
       "2          Precision  0.719225    0.714286   -0.004940          Precision   \n",
       "3             Recall  0.648791    0.729167    0.080376             Recall   \n",
       "4           F1 Score  0.677118    0.721649    0.044532           F1 Score   \n",
       "5       Fit Time (s)  1.783538    4.905923    3.122385       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.669984    0.728220    0.058236  \n",
       "1  0.739126    0.792803    0.053677  \n",
       "2  0.679976    0.700000    0.020024  \n",
       "3  0.642880    0.729167    0.086286  \n",
       "4  0.650457    0.714286    0.063829  \n",
       "5  1.660865    6.081828    4.420963  "
      ]
     },
     "execution_count": 118,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_rf_roi_gs, result_df_rf_roi_gs_groups], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "RF - VBM\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 9 candidates, totalling 45 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 924.644660949707\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / VBM / GridSearchCV / StratKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_vbm_gs = evaluate_model_gridsearch(\n",
    "    model=model_rf,\n",
    "    param_grid=param_grid_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.669979</td>\n",
       "      <td>0.75523</td>\n",
       "      <td>0.686865</td>\n",
       "      <td>0.568848</td>\n",
       "      <td>0.61693</td>\n",
       "      <td>15.152294</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.599242</td>\n",
       "      <td>0.691098</td>\n",
       "      <td>0.625</td>\n",
       "      <td>0.416667</td>\n",
       "      <td>0.5</td>\n",
       "      <td>206.325187</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>-0.070737</td>\n",
       "      <td>-0.064131</td>\n",
       "      <td>-0.061865</td>\n",
       "      <td>-0.152181</td>\n",
       "      <td>-0.11693</td>\n",
       "      <td>191.172894</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.669979   0.75523   0.686865  0.568848   0.61693   \n",
       "GridSearch           0.599242  0.691098      0.625  0.416667       0.5   \n",
       "Difference          -0.070737 -0.064131  -0.061865 -0.152181  -0.11693   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline       15.152294  \n",
       "GridSearch    206.325187  \n",
       "Difference    191.172894  "
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_vbm_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 9 candidates, totalling 27 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 391.14190196990967\n"
     ]
    }
   ],
   "source": [
    "# ---------- RandomForest / VBM / GridSearchCV / StratGroupKFold\n",
    "\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_rf_vbm_gs_groups = evaluate_model_gridsearch(\n",
    "    model=model_rf,\n",
    "    param_grid=param_grid_rf,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.659715</td>\n",
       "      <td>0.585038</td>\n",
       "      <td>-0.074677</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.733362</td>\n",
       "      <td>0.678788</td>\n",
       "      <td>-0.054574</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.670235</td>\n",
       "      <td>0.575000</td>\n",
       "      <td>-0.095235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.575758</td>\n",
       "      <td>0.479167</td>\n",
       "      <td>-0.096591</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.609792</td>\n",
       "      <td>0.522727</td>\n",
       "      <td>-0.087064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>12.383874</td>\n",
       "      <td>118.558576</td>\n",
       "      <td>106.174702</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric   Baseline  GridSearch  Difference\n",
       "0  Balanced Accuracy   0.659715    0.585038   -0.074677\n",
       "1            ROC-AUC   0.733362    0.678788   -0.054574\n",
       "2          Precision   0.670235    0.575000   -0.095235\n",
       "3             Recall   0.575758    0.479167   -0.096591\n",
       "4           F1 Score   0.609792    0.522727   -0.087064\n",
       "5       Fit Time (s)  12.383874  118.558576  106.174702"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_rf_vbm_gs_groups\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.669979</td>\n",
       "      <td>0.599242</td>\n",
       "      <td>-0.070737</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.659715</td>\n",
       "      <td>0.585038</td>\n",
       "      <td>-0.074677</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.755230</td>\n",
       "      <td>0.691098</td>\n",
       "      <td>-0.064131</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.733362</td>\n",
       "      <td>0.678788</td>\n",
       "      <td>-0.054574</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.686865</td>\n",
       "      <td>0.625000</td>\n",
       "      <td>-0.061865</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.670235</td>\n",
       "      <td>0.575000</td>\n",
       "      <td>-0.095235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.568848</td>\n",
       "      <td>0.416667</td>\n",
       "      <td>-0.152181</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.575758</td>\n",
       "      <td>0.479167</td>\n",
       "      <td>-0.096591</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.616930</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>-0.116930</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.609792</td>\n",
       "      <td>0.522727</td>\n",
       "      <td>-0.087064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>15.152294</td>\n",
       "      <td>206.325187</td>\n",
       "      <td>191.172894</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>12.383874</td>\n",
       "      <td>118.558576</td>\n",
       "      <td>106.174702</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric   Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy   0.669979    0.599242   -0.070737  Balanced Accuracy   \n",
       "1            ROC-AUC   0.755230    0.691098   -0.064131            ROC-AUC   \n",
       "2          Precision   0.686865    0.625000   -0.061865          Precision   \n",
       "3             Recall   0.568848    0.416667   -0.152181             Recall   \n",
       "4           F1 Score   0.616930    0.500000   -0.116930           F1 Score   \n",
       "5       Fit Time (s)  15.152294  206.325187  191.172894       Fit Time (s)   \n",
       "\n",
       "    Baseline  GridSearch  Difference  \n",
       "0   0.659715    0.585038   -0.074677  \n",
       "1   0.733362    0.678788   -0.054574  \n",
       "2   0.670235    0.575000   -0.095235  \n",
       "3   0.575758    0.479167   -0.096591  \n",
       "4   0.609792    0.522727   -0.087064  \n",
       "5  12.383874  118.558576  106.174702  "
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_rf_vbm_gs, result_df_rf_vbm_gs_groups], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "XGBBoost\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import VotingClassifier\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.svm import SVC\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- XGBoost / Definitions\n",
    "\n",
    "model_xgboost = XGBClassifier()\n",
    "\n",
    "\n",
    "param_grid_xgboost = {\n",
    "    \"xgbclassifier__n_estimators\": [100, 200, 300],  # Parameters for XGBClassifier\n",
    "    \"xgbclassifier__learning_rate\": [0.1, 0.01, 0.001],\n",
    "    \"xgbclassifier__max_depth\": [3, 5, 7],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 27 candidates, totalling 135 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "python(17252) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17253) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "python(17254) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17255) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17256) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17257) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17258) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17259) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17260) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17261) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17262) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(17263) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 111.87668776512146\n"
     ]
    }
   ],
   "source": [
    "# ---------- XGBoost / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_xgb_roi_gs = evaluate_model_gridsearch(\n",
    "    model=model_xgboost,\n",
    "    param_grid=param_grid_xgboost,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.734116</td>\n",
       "      <td>0.808504</td>\n",
       "      <td>0.738544</td>\n",
       "      <td>0.670555</td>\n",
       "      <td>0.698943</td>\n",
       "      <td>1.615379</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.703409</td>\n",
       "      <td>0.82803</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>0.625</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>8.205869</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>-0.030707</td>\n",
       "      <td>0.019527</td>\n",
       "      <td>-0.024258</td>\n",
       "      <td>-0.045555</td>\n",
       "      <td>-0.032276</td>\n",
       "      <td>6.59049</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.734116  0.808504   0.738544  0.670555  0.698943   \n",
       "GridSearch           0.703409   0.82803   0.714286     0.625  0.666667   \n",
       "Difference          -0.030707  0.019527  -0.024258 -0.045555 -0.032276   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.615379  \n",
       "GridSearch      8.205869  \n",
       "Difference       6.59049  "
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_xgb_roi_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 27 candidates, totalling 81 fits\n",
      "Elapsed time: 59.216166973114014\n"
     ]
    }
   ],
   "source": [
    "# ---------- XGBoost / ROI / GridSearchCV / StratGroupKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_xgb_roi_gs_groups = evaluate_model_gridsearch(\n",
    "    model=model_xgboost,\n",
    "    param_grid=param_grid_xgboost,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.720797</td>\n",
       "      <td>0.791891</td>\n",
       "      <td>0.70579</td>\n",
       "      <td>0.715054</td>\n",
       "      <td>0.709294</td>\n",
       "      <td>2.239068</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.754167</td>\n",
       "      <td>0.82803</td>\n",
       "      <td>0.755556</td>\n",
       "      <td>0.708333</td>\n",
       "      <td>0.731183</td>\n",
       "      <td>6.874486</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.033369</td>\n",
       "      <td>0.036139</td>\n",
       "      <td>0.049766</td>\n",
       "      <td>-0.00672</td>\n",
       "      <td>0.021889</td>\n",
       "      <td>4.635418</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.720797  0.791891    0.70579  0.715054  0.709294   \n",
       "GridSearch           0.754167   0.82803   0.755556  0.708333  0.731183   \n",
       "Difference           0.033369  0.036139   0.049766  -0.00672  0.021889   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        2.239068  \n",
       "GridSearch      6.874486  \n",
       "Difference      4.635418  "
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_xgb_roi_gs_groups.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.734116</td>\n",
       "      <td>0.703409</td>\n",
       "      <td>-0.030707</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.720797</td>\n",
       "      <td>0.754167</td>\n",
       "      <td>0.033369</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.808504</td>\n",
       "      <td>0.828030</td>\n",
       "      <td>0.019527</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.791891</td>\n",
       "      <td>0.828030</td>\n",
       "      <td>0.036139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.738544</td>\n",
       "      <td>0.714286</td>\n",
       "      <td>-0.024258</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.705790</td>\n",
       "      <td>0.755556</td>\n",
       "      <td>0.049766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.670555</td>\n",
       "      <td>0.625000</td>\n",
       "      <td>-0.045555</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.715054</td>\n",
       "      <td>0.708333</td>\n",
       "      <td>-0.006720</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.698943</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>-0.032276</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.709294</td>\n",
       "      <td>0.731183</td>\n",
       "      <td>0.021889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.615379</td>\n",
       "      <td>8.205869</td>\n",
       "      <td>6.590490</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>2.239068</td>\n",
       "      <td>6.874486</td>\n",
       "      <td>4.635418</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.734116    0.703409   -0.030707  Balanced Accuracy   \n",
       "1            ROC-AUC  0.808504    0.828030    0.019527            ROC-AUC   \n",
       "2          Precision  0.738544    0.714286   -0.024258          Precision   \n",
       "3             Recall  0.670555    0.625000   -0.045555             Recall   \n",
       "4           F1 Score  0.698943    0.666667   -0.032276           F1 Score   \n",
       "5       Fit Time (s)  1.615379    8.205869    6.590490       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.720797    0.754167    0.033369  \n",
       "1  0.791891    0.828030    0.036139  \n",
       "2  0.705790    0.755556    0.049766  \n",
       "3  0.715054    0.708333   -0.006720  \n",
       "4  0.709294    0.731183    0.021889  \n",
       "5  2.239068    6.874486    4.635418  "
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_xgb_roi_gs, result_df_xgb_roi_gs_groups], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "KNN\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- KNN / Definitions\n",
    "\n",
    "# Define the model\n",
    "model = KNeighborsClassifier()\n",
    "\n",
    "# Define the parameter grid for GridSearchCV\n",
    "param_grid = {\n",
    "    \"kneighborsclassifier__n_neighbors\": [3, 5, 7, 9],\n",
    "    \"kneighborsclassifier__weights\": [\"uniform\", \"distance\"],\n",
    "    \"kneighborsclassifier__metric\": [\"euclidean\", \"manhattan\"],\n",
    "}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 16 candidates, totalling 80 fits\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Elapsed time: 55.49656701087952\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df = evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,  # Pass groups if available, otherwise remove this argument\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.694637</td>\n",
       "      <td>0.765214</td>\n",
       "      <td>0.737511</td>\n",
       "      <td>0.569275</td>\n",
       "      <td>0.638216</td>\n",
       "      <td>1.424062</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.724242</td>\n",
       "      <td>0.780303</td>\n",
       "      <td>0.727273</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.695652</td>\n",
       "      <td>6.723507</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.029605</td>\n",
       "      <td>0.015089</td>\n",
       "      <td>-0.010238</td>\n",
       "      <td>0.097392</td>\n",
       "      <td>0.057436</td>\n",
       "      <td>5.299445</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.694637  0.765214   0.737511  0.569275  0.638216   \n",
       "GridSearch           0.724242  0.780303   0.727273  0.666667  0.695652   \n",
       "Difference           0.029605  0.015089  -0.010238  0.097392  0.057436   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.424062  \n",
       "GridSearch      6.723507  \n",
       "Difference      5.299445  "
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_roi_gs = result_df.copy()\n",
    "result_df_knn_roi_gs.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 16 candidates, totalling 48 fits\n",
      "Elapsed time: 27.71569800376892\n"
     ]
    }
   ],
   "source": [
    "# ---------- KNN / ROI / GridSearchCV / StratGroupKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_roi_gs_groups = evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Metric</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>Precision</td>\n",
       "      <td>Recall</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Baseline</th>\n",
       "      <td>0.614426</td>\n",
       "      <td>0.643103</td>\n",
       "      <td>0.617445</td>\n",
       "      <td>0.583252</td>\n",
       "      <td>0.592539</td>\n",
       "      <td>1.784181</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GridSearch</th>\n",
       "      <td>0.686553</td>\n",
       "      <td>0.745076</td>\n",
       "      <td>0.673913</td>\n",
       "      <td>0.645833</td>\n",
       "      <td>0.659574</td>\n",
       "      <td>5.038332</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Difference</th>\n",
       "      <td>0.072127</td>\n",
       "      <td>0.101972</td>\n",
       "      <td>0.056468</td>\n",
       "      <td>0.062581</td>\n",
       "      <td>0.067036</td>\n",
       "      <td>3.254151</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                            0         1          2         3         4  \\\n",
       "Metric      Balanced Accuracy   ROC-AUC  Precision    Recall  F1 Score   \n",
       "Baseline             0.614426  0.643103   0.617445  0.583252  0.592539   \n",
       "GridSearch           0.686553  0.745076   0.673913  0.645833  0.659574   \n",
       "Difference           0.072127  0.101972   0.056468  0.062581  0.067036   \n",
       "\n",
       "                       5  \n",
       "Metric      Fit Time (s)  \n",
       "Baseline        1.784181  \n",
       "GridSearch      5.038332  \n",
       "Difference      3.254151  "
      ]
     },
     "execution_count": 140,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_knn_roi_gs_groups.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.694637</td>\n",
       "      <td>0.724242</td>\n",
       "      <td>0.029605</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.614426</td>\n",
       "      <td>0.686553</td>\n",
       "      <td>0.072127</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.765214</td>\n",
       "      <td>0.780303</td>\n",
       "      <td>0.015089</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.643103</td>\n",
       "      <td>0.745076</td>\n",
       "      <td>0.101972</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.737511</td>\n",
       "      <td>0.727273</td>\n",
       "      <td>-0.010238</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.617445</td>\n",
       "      <td>0.673913</td>\n",
       "      <td>0.056468</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.569275</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.097392</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.583252</td>\n",
       "      <td>0.645833</td>\n",
       "      <td>0.062581</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.638216</td>\n",
       "      <td>0.695652</td>\n",
       "      <td>0.057436</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.592539</td>\n",
       "      <td>0.659574</td>\n",
       "      <td>0.067036</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.424062</td>\n",
       "      <td>6.723507</td>\n",
       "      <td>5.299445</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>1.784181</td>\n",
       "      <td>5.038332</td>\n",
       "      <td>3.254151</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.694637    0.724242    0.029605  Balanced Accuracy   \n",
       "1            ROC-AUC  0.765214    0.780303    0.015089            ROC-AUC   \n",
       "2          Precision  0.737511    0.727273   -0.010238          Precision   \n",
       "3             Recall  0.569275    0.666667    0.097392             Recall   \n",
       "4           F1 Score  0.638216    0.695652    0.057436           F1 Score   \n",
       "5       Fit Time (s)  1.424062    6.723507    5.299445       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.614426    0.686553    0.072127  \n",
       "1  0.643103    0.745076    0.101972  \n",
       "2  0.617445    0.673913    0.056468  \n",
       "3  0.583252    0.645833    0.062581  \n",
       "4  0.592539    0.659574    0.067036  \n",
       "5  1.784181    5.038332    3.254151  "
      ]
     },
     "execution_count": 141,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_knn_roi_gs, result_df_knn_roi_gs_groups], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "KNN - VBM\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 16 candidates, totalling 80 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "python(22246) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22247) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22248) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22249) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22250) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22251) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22252) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22253) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22254) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22255) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22256) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n",
      "python(22257) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[152], line 7\u001b[0m\n\u001b[1;32m      4\u001b[0m f_extractor \u001b[38;5;241m=\u001b[39m VBMFeatureExtractor()\n\u001b[1;32m      6\u001b[0m start_time \u001b[38;5;241m=\u001b[39m time\u001b[38;5;241m.\u001b[39mtime()\n\u001b[0;32m----> 7\u001b[0m result_df_knn_vbm \u001b[38;5;241m=\u001b[39m \u001b[43mevaluate_model_gridsearch\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m      8\u001b[0m \u001b[43m    \u001b[49m\u001b[43mmodel\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      9\u001b[0m \u001b[43m    \u001b[49m\u001b[43mparam_grid\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     10\u001b[0m \u001b[43m    \u001b[49m\u001b[43mf_extractor\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     11\u001b[0m \u001b[43m    \u001b[49m\u001b[43mX_train\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     12\u001b[0m \u001b[43m    \u001b[49m\u001b[43my_train_encoded\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     13\u001b[0m \u001b[43m    \u001b[49m\u001b[43mX_test\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     14\u001b[0m \u001b[43m    \u001b[49m\u001b[43my_test_encoded\u001b[49m\u001b[43m,\u001b[49m\u001b[43m  \u001b[49m\u001b[38;5;66;43;03m# Pass groups if available, otherwise remove this argument\u001b[39;49;00m\n\u001b[1;32m     15\u001b[0m \u001b[43m)\u001b[49m\n\u001b[1;32m     16\u001b[0m end_time \u001b[38;5;241m=\u001b[39m time\u001b[38;5;241m.\u001b[39mtime()\n\u001b[1;32m     17\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mElapsed time: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mend_time\u001b[38;5;241m-\u001b[39mstart_time\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m\"\u001b[39m)\n",
      "Cell \u001b[0;32mIn[106], line 31\u001b[0m, in \u001b[0;36mevaluate_model_gridsearch\u001b[0;34m(model, param_grid, f_extractor, X_train, y_train, X_test, y_test, groups)\u001b[0m\n\u001b[1;32m     20\u001b[0m grid_search \u001b[38;5;241m=\u001b[39m GridSearchCV(\n\u001b[1;32m     21\u001b[0m     estimator,\n\u001b[1;32m     22\u001b[0m     param_grid,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m     27\u001b[0m     n_jobs\u001b[38;5;241m=\u001b[39m\u001b[38;5;241m-\u001b[39m\u001b[38;5;241m1\u001b[39m,\n\u001b[1;32m     28\u001b[0m )\n\u001b[1;32m     30\u001b[0m \u001b[38;5;66;03m# Fitting GridSearchCV to find the best parameters and model, and measure fitting time\u001b[39;00m\n\u001b[0;32m---> 31\u001b[0m \u001b[43mgrid_search\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfit\u001b[49m\u001b[43m(\u001b[49m\u001b[43mX_train\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43my_train\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mgroups\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mgroups\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     32\u001b[0m best_model \u001b[38;5;241m=\u001b[39m grid_search\u001b[38;5;241m.\u001b[39mbest_estimator_\n\u001b[1;32m     33\u001b[0m grid_search_fit_time \u001b[38;5;241m=\u001b[39m grid_search\u001b[38;5;241m.\u001b[39mcv_results_[\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mmean_fit_time\u001b[39m\u001b[38;5;124m\"\u001b[39m]\u001b[38;5;241m.\u001b[39mmean()\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/sklearn/base.py:1474\u001b[0m, in \u001b[0;36m_fit_context.<locals>.decorator.<locals>.wrapper\u001b[0;34m(estimator, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1467\u001b[0m     estimator\u001b[38;5;241m.\u001b[39m_validate_params()\n\u001b[1;32m   1469\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m config_context(\n\u001b[1;32m   1470\u001b[0m     skip_parameter_validation\u001b[38;5;241m=\u001b[39m(\n\u001b[1;32m   1471\u001b[0m         prefer_skip_nested_validation \u001b[38;5;129;01mor\u001b[39;00m global_skip_validation\n\u001b[1;32m   1472\u001b[0m     )\n\u001b[1;32m   1473\u001b[0m ):\n\u001b[0;32m-> 1474\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mfit_method\u001b[49m\u001b[43m(\u001b[49m\u001b[43mestimator\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/sklearn/model_selection/_search.py:970\u001b[0m, in \u001b[0;36mBaseSearchCV.fit\u001b[0;34m(self, X, y, **params)\u001b[0m\n\u001b[1;32m    964\u001b[0m     results \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_format_results(\n\u001b[1;32m    965\u001b[0m         all_candidate_params, n_splits, all_out, all_more_results\n\u001b[1;32m    966\u001b[0m     )\n\u001b[1;32m    968\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m results\n\u001b[0;32m--> 970\u001b[0m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_run_search\u001b[49m\u001b[43m(\u001b[49m\u001b[43mevaluate_candidates\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    972\u001b[0m \u001b[38;5;66;03m# multimetric is determined here because in the case of a callable\u001b[39;00m\n\u001b[1;32m    973\u001b[0m \u001b[38;5;66;03m# self.scoring the return type is only known after calling\u001b[39;00m\n\u001b[1;32m    974\u001b[0m first_test_score \u001b[38;5;241m=\u001b[39m all_out[\u001b[38;5;241m0\u001b[39m][\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mtest_scores\u001b[39m\u001b[38;5;124m\"\u001b[39m]\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/sklearn/model_selection/_search.py:1527\u001b[0m, in \u001b[0;36mGridSearchCV._run_search\u001b[0;34m(self, evaluate_candidates)\u001b[0m\n\u001b[1;32m   1525\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m_run_search\u001b[39m(\u001b[38;5;28mself\u001b[39m, evaluate_candidates):\n\u001b[1;32m   1526\u001b[0m \u001b[38;5;250m    \u001b[39m\u001b[38;5;124;03m\"\"\"Search all candidates in param_grid\"\"\"\u001b[39;00m\n\u001b[0;32m-> 1527\u001b[0m     \u001b[43mevaluate_candidates\u001b[49m\u001b[43m(\u001b[49m\u001b[43mParameterGrid\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mparam_grid\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/sklearn/model_selection/_search.py:916\u001b[0m, in \u001b[0;36mBaseSearchCV.fit.<locals>.evaluate_candidates\u001b[0;34m(candidate_params, cv, more_results)\u001b[0m\n\u001b[1;32m    908\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mverbose \u001b[38;5;241m>\u001b[39m \u001b[38;5;241m0\u001b[39m:\n\u001b[1;32m    909\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\n\u001b[1;32m    910\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mFitting \u001b[39m\u001b[38;5;132;01m{0}\u001b[39;00m\u001b[38;5;124m folds for each of \u001b[39m\u001b[38;5;132;01m{1}\u001b[39;00m\u001b[38;5;124m candidates,\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    911\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m totalling \u001b[39m\u001b[38;5;132;01m{2}\u001b[39;00m\u001b[38;5;124m fits\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m.\u001b[39mformat(\n\u001b[1;32m    912\u001b[0m             n_splits, n_candidates, n_candidates \u001b[38;5;241m*\u001b[39m n_splits\n\u001b[1;32m    913\u001b[0m         )\n\u001b[1;32m    914\u001b[0m     )\n\u001b[0;32m--> 916\u001b[0m out \u001b[38;5;241m=\u001b[39m \u001b[43mparallel\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    917\u001b[0m \u001b[43m    \u001b[49m\u001b[43mdelayed\u001b[49m\u001b[43m(\u001b[49m\u001b[43m_fit_and_score\u001b[49m\u001b[43m)\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    918\u001b[0m \u001b[43m        \u001b[49m\u001b[43mclone\u001b[49m\u001b[43m(\u001b[49m\u001b[43mbase_estimator\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    919\u001b[0m \u001b[43m        \u001b[49m\u001b[43mX\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    920\u001b[0m \u001b[43m        \u001b[49m\u001b[43my\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    921\u001b[0m \u001b[43m        \u001b[49m\u001b[43mtrain\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtrain\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    922\u001b[0m \u001b[43m        \u001b[49m\u001b[43mtest\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtest\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    923\u001b[0m \u001b[43m        \u001b[49m\u001b[43mparameters\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mparameters\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    924\u001b[0m \u001b[43m        \u001b[49m\u001b[43msplit_progress\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43msplit_idx\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mn_splits\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    925\u001b[0m \u001b[43m        \u001b[49m\u001b[43mcandidate_progress\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43mcand_idx\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mn_candidates\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    926\u001b[0m \u001b[43m        \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mfit_and_score_kwargs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    927\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    928\u001b[0m \u001b[43m    \u001b[49m\u001b[38;5;28;43;01mfor\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43m(\u001b[49m\u001b[43mcand_idx\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mparameters\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m(\u001b[49m\u001b[43msplit_idx\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43m(\u001b[49m\u001b[43mtrain\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mtest\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;129;43;01min\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mproduct\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    929\u001b[0m \u001b[43m        \u001b[49m\u001b[38;5;28;43menumerate\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43mcandidate_params\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    930\u001b[0m \u001b[43m        \u001b[49m\u001b[38;5;28;43menumerate\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43mcv\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msplit\u001b[49m\u001b[43m(\u001b[49m\u001b[43mX\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43my\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mrouted_params\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msplitter\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msplit\u001b[49m\u001b[43m)\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    931\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    932\u001b[0m \u001b[43m\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    934\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mlen\u001b[39m(out) \u001b[38;5;241m<\u001b[39m \u001b[38;5;241m1\u001b[39m:\n\u001b[1;32m    935\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(\n\u001b[1;32m    936\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mNo fits were performed. \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    937\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mWas the CV iterator empty? \u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    938\u001b[0m         \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mWere there no candidates?\u001b[39m\u001b[38;5;124m\"\u001b[39m\n\u001b[1;32m    939\u001b[0m     )\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/sklearn/utils/parallel.py:67\u001b[0m, in \u001b[0;36mParallel.__call__\u001b[0;34m(self, iterable)\u001b[0m\n\u001b[1;32m     62\u001b[0m config \u001b[38;5;241m=\u001b[39m get_config()\n\u001b[1;32m     63\u001b[0m iterable_with_config \u001b[38;5;241m=\u001b[39m (\n\u001b[1;32m     64\u001b[0m     (_with_config(delayed_func, config), args, kwargs)\n\u001b[1;32m     65\u001b[0m     \u001b[38;5;28;01mfor\u001b[39;00m delayed_func, args, kwargs \u001b[38;5;129;01min\u001b[39;00m iterable\n\u001b[1;32m     66\u001b[0m )\n\u001b[0;32m---> 67\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43msuper\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[38;5;21;43m__call__\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43miterable_with_config\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/joblib/parallel.py:1952\u001b[0m, in \u001b[0;36mParallel.__call__\u001b[0;34m(self, iterable)\u001b[0m\n\u001b[1;32m   1946\u001b[0m \u001b[38;5;66;03m# The first item from the output is blank, but it makes the interpreter\u001b[39;00m\n\u001b[1;32m   1947\u001b[0m \u001b[38;5;66;03m# progress until it enters the Try/Except block of the generator and\u001b[39;00m\n\u001b[1;32m   1948\u001b[0m \u001b[38;5;66;03m# reach the first `yield` statement. This starts the aynchronous\u001b[39;00m\n\u001b[1;32m   1949\u001b[0m \u001b[38;5;66;03m# dispatch of the tasks to the workers.\u001b[39;00m\n\u001b[1;32m   1950\u001b[0m \u001b[38;5;28mnext\u001b[39m(output)\n\u001b[0;32m-> 1952\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m output \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mreturn_generator \u001b[38;5;28;01melse\u001b[39;00m \u001b[38;5;28;43mlist\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43moutput\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/joblib/parallel.py:1595\u001b[0m, in \u001b[0;36mParallel._get_outputs\u001b[0;34m(self, iterator, pre_dispatch)\u001b[0m\n\u001b[1;32m   1592\u001b[0m     \u001b[38;5;28;01myield\u001b[39;00m\n\u001b[1;32m   1594\u001b[0m     \u001b[38;5;28;01mwith\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backend\u001b[38;5;241m.\u001b[39mretrieval_context():\n\u001b[0;32m-> 1595\u001b[0m         \u001b[38;5;28;01myield from\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_retrieve()\n\u001b[1;32m   1597\u001b[0m \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mGeneratorExit\u001b[39;00m:\n\u001b[1;32m   1598\u001b[0m     \u001b[38;5;66;03m# The generator has been garbage collected before being fully\u001b[39;00m\n\u001b[1;32m   1599\u001b[0m     \u001b[38;5;66;03m# consumed. This aborts the remaining tasks if possible and warn\u001b[39;00m\n\u001b[1;32m   1600\u001b[0m     \u001b[38;5;66;03m# the user if necessary.\u001b[39;00m\n\u001b[1;32m   1601\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_exception \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mTrue\u001b[39;00m\n",
      "File \u001b[0;32m~/anaconda3/envs/brain-anatomy-schizophrenia/lib/python3.12/site-packages/joblib/parallel.py:1707\u001b[0m, in \u001b[0;36mParallel._retrieve\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1702\u001b[0m \u001b[38;5;66;03m# If the next job is not ready for retrieval yet, we just wait for\u001b[39;00m\n\u001b[1;32m   1703\u001b[0m \u001b[38;5;66;03m# async callbacks to progress.\u001b[39;00m\n\u001b[1;32m   1704\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m ((\u001b[38;5;28mlen\u001b[39m(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_jobs) \u001b[38;5;241m==\u001b[39m \u001b[38;5;241m0\u001b[39m) \u001b[38;5;129;01mor\u001b[39;00m\n\u001b[1;32m   1705\u001b[0m     (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_jobs[\u001b[38;5;241m0\u001b[39m]\u001b[38;5;241m.\u001b[39mget_status(\n\u001b[1;32m   1706\u001b[0m         timeout\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mtimeout) \u001b[38;5;241m==\u001b[39m TASK_PENDING)):\n\u001b[0;32m-> 1707\u001b[0m     \u001b[43mtime\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msleep\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m0.01\u001b[39;49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1708\u001b[0m     \u001b[38;5;28;01mcontinue\u001b[39;00m\n\u001b[1;32m   1710\u001b[0m \u001b[38;5;66;03m# We need to be careful: the job list can be filling up as\u001b[39;00m\n\u001b[1;32m   1711\u001b[0m \u001b[38;5;66;03m# we empty it and Python list are not thread-safe by\u001b[39;00m\n\u001b[1;32m   1712\u001b[0m \u001b[38;5;66;03m# default hence the use of the lock\u001b[39;00m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "# ---------- KNN / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_vbm = evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,  # Pass groups if available, otherwise remove this argument\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "result_df_knn_vbm.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- KNN / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "# Define the feature extractor\n",
    "f_extractor = VBMFeatureExtractor()\n",
    "\n",
    "start_time = time.time()\n",
    "result_df_knn_vbm_group = evaluate_model_gridsearch(\n",
    "    model,\n",
    "    param_grid,\n",
    "    f_extractor,\n",
    "    X_train,\n",
    "    y_train_encoded,\n",
    "    X_test,\n",
    "    y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "result_df_knn_vbm_group.T\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "megatable = pd.concat([result_df_knn_vbm, result_df_knn_vbm_group], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVC, kernel='rbf'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ---------- SVC, rbf kernel / Definitions\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from sklearn.model_selection import GridSearchCV, cross_validate\n",
    "from sklearn.metrics import balanced_accuracy_score, roc_auc_score\n",
    "\n",
    "\n",
    "model_svc = SVC(kernel=\"rbf\", probability=True)\n",
    "\n",
    "# Define the parameter grid for the SVC\n",
    "param_grid_svc = {\"svc__C\": [0.1, 1, 10], \"svc__gamma\": [1, 0.1, 0.01]}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 9 candidates, totalling 45 fits\n",
      "Elapsed time: 24.61484694480896\n"
     ]
    }
   ],
   "source": [
    "# ---------- SVC, rbf kernel / ROI / GridSearchCV / StratKFold\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "# Call the evaluate_model_gridsearch function with the SVC model and parameter grid\n",
    "start_time = time.time()\n",
    "result_df_svc_rbf_gs = evaluate_model_gridsearch(\n",
    "    model=model_svc,\n",
    "    param_grid=param_grid_svc,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.632443</td>\n",
       "      <td>0.700947</td>\n",
       "      <td>0.068504</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.699470</td>\n",
       "      <td>0.725758</td>\n",
       "      <td>0.026288</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.655911</td>\n",
       "      <td>0.660377</td>\n",
       "      <td>0.004466</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.499431</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.229736</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.562842</td>\n",
       "      <td>0.693069</td>\n",
       "      <td>0.130228</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>0.653082</td>\n",
       "      <td>4.976191</td>\n",
       "      <td>4.323109</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference\n",
       "0  Balanced Accuracy  0.632443    0.700947    0.068504\n",
       "1            ROC-AUC  0.699470    0.725758    0.026288\n",
       "2          Precision  0.655911    0.660377    0.004466\n",
       "3             Recall  0.499431    0.729167    0.229736\n",
       "4           F1 Score  0.562842    0.693069    0.130228\n",
       "5       Fit Time (s)  0.653082    4.976191    4.323109"
      ]
     },
     "execution_count": 178,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svc_rbf_gs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 2 folds for each of 9 candidates, totalling 18 fits\n",
      "Elapsed time: 16.476353883743286\n"
     ]
    }
   ],
   "source": [
    "# SVC RBF / ROI / Group\n",
    "\n",
    "f_extractor = ROIsFeatureExtractor()\n",
    "\n",
    "# Call the evaluate_model_gridsearch function with the SVC model and parameter grid\n",
    "start_time = time.time()\n",
    "result_df_svc_rbf_gs_group = evaluate_model_gridsearch(\n",
    "    model=model_svc,\n",
    "    param_grid=param_grid_svc,\n",
    "    f_extractor=f_extractor,\n",
    "    X_train=X_train,\n",
    "    y_train=y_train_encoded,\n",
    "    X_test=X_test,\n",
    "    y_test=y_test_encoded,\n",
    "    groups=groups,\n",
    ")\n",
    "end_time = time.time()\n",
    "print(f\"Elapsed time: {end_time-start_time}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.589176</td>\n",
       "      <td>0.651515</td>\n",
       "      <td>0.062339</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.659538</td>\n",
       "      <td>0.739394</td>\n",
       "      <td>0.079856</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.617262</td>\n",
       "      <td>0.615385</td>\n",
       "      <td>-0.001878</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.514827</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.151839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.542571</td>\n",
       "      <td>0.640000</td>\n",
       "      <td>0.097429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>2.109760</td>\n",
       "      <td>7.005453</td>\n",
       "      <td>4.895693</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference\n",
       "0  Balanced Accuracy  0.589176    0.651515    0.062339\n",
       "1            ROC-AUC  0.659538    0.739394    0.079856\n",
       "2          Precision  0.617262    0.615385   -0.001878\n",
       "3             Recall  0.514827    0.666667    0.151839\n",
       "4           F1 Score  0.542571    0.640000    0.097429\n",
       "5       Fit Time (s)  2.109760    7.005453    4.895693"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "result_df_svc_rbf_gs_group\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "      <th>Metric</th>\n",
       "      <th>Baseline</th>\n",
       "      <th>GridSearch</th>\n",
       "      <th>Difference</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.632443</td>\n",
       "      <td>0.700947</td>\n",
       "      <td>0.068504</td>\n",
       "      <td>Balanced Accuracy</td>\n",
       "      <td>0.589176</td>\n",
       "      <td>0.651515</td>\n",
       "      <td>0.062339</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.699470</td>\n",
       "      <td>0.725758</td>\n",
       "      <td>0.026288</td>\n",
       "      <td>ROC-AUC</td>\n",
       "      <td>0.659538</td>\n",
       "      <td>0.739394</td>\n",
       "      <td>0.079856</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Precision</td>\n",
       "      <td>0.655911</td>\n",
       "      <td>0.660377</td>\n",
       "      <td>0.004466</td>\n",
       "      <td>Precision</td>\n",
       "      <td>0.617262</td>\n",
       "      <td>0.615385</td>\n",
       "      <td>-0.001878</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Recall</td>\n",
       "      <td>0.499431</td>\n",
       "      <td>0.729167</td>\n",
       "      <td>0.229736</td>\n",
       "      <td>Recall</td>\n",
       "      <td>0.514827</td>\n",
       "      <td>0.666667</td>\n",
       "      <td>0.151839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.562842</td>\n",
       "      <td>0.693069</td>\n",
       "      <td>0.130228</td>\n",
       "      <td>F1 Score</td>\n",
       "      <td>0.542571</td>\n",
       "      <td>0.640000</td>\n",
       "      <td>0.097429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>0.653082</td>\n",
       "      <td>4.976191</td>\n",
       "      <td>4.323109</td>\n",
       "      <td>Fit Time (s)</td>\n",
       "      <td>2.109760</td>\n",
       "      <td>7.005453</td>\n",
       "      <td>4.895693</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "              Metric  Baseline  GridSearch  Difference             Metric  \\\n",
       "0  Balanced Accuracy  0.632443    0.700947    0.068504  Balanced Accuracy   \n",
       "1            ROC-AUC  0.699470    0.725758    0.026288            ROC-AUC   \n",
       "2          Precision  0.655911    0.660377    0.004466          Precision   \n",
       "3             Recall  0.499431    0.729167    0.229736             Recall   \n",
       "4           F1 Score  0.562842    0.693069    0.130228           F1 Score   \n",
       "5       Fit Time (s)  0.653082    4.976191    4.323109       Fit Time (s)   \n",
       "\n",
       "   Baseline  GridSearch  Difference  \n",
       "0  0.589176    0.651515    0.062339  \n",
       "1  0.659538    0.739394    0.079856  \n",
       "2  0.617262    0.615385   -0.001878  \n",
       "3  0.514827    0.666667    0.151839  \n",
       "4  0.542571    0.640000    0.097429  \n",
       "5  2.109760    7.005453    4.895693  "
      ]
     },
     "execution_count": 181,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "megatable = pd.concat([result_df_svc_rbf_gs, result_df_svc_rbf_gs_group], axis=1)\n",
    "megatable\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# III. Test submission locally\n",
    "\n",
    "The submissions need to be located in the submissions folder. For instance for `starting_kit`, it should be located in `submissions/submissions/starting_kit`.\n",
    "\n",
    "Copy everything required (the cell bellow) to build your estimator in a submission file: `submissions/submissions/starting_kit/estimator.py`. This file must contain a function `get_estimator()`:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!/usr/bin/env python3\n",
    "# -*- coding: utf-8 -*-\n",
    "\"\"\"\n",
    "Created on Wed Nov 30 02:11:17 2022\n",
    "\n",
    "@author: edouard.duchesnay@cea.fr\n",
    "\"\"\"\n",
    "\n",
    "# import numpy as np\n",
    "\n",
    "from sklearn.base import BaseEstimator\n",
    "from sklearn.base import TransformerMixin\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.pipeline import make_pipeline\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "\n",
    "class ROIsFeatureExtractor(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"Select only the 284 ROIs features:\"\"\"\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        return self\n",
    "\n",
    "    def transform(self, X):\n",
    "        return X[:, :284]\n",
    "\n",
    "\n",
    "def get_estimator():\n",
    "    \"\"\"Build your estimator here.\"\"\"\n",
    "    estimator = make_pipeline(ROIsFeatureExtractor(), XGBClassifier())\n",
    "\n",
    "    return estimator\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!/usr/bin/env python3\n",
    "# -*- coding: utf-8 -*-\n",
    "\"\"\"\n",
    "Created on Wed Nov 30 02:11:17 2022\n",
    "\n",
    "@author: edouard.duchesnay@cea.fr\n",
    "\"\"\"\n",
    "\n",
    "# import numpy as np\n",
    "\n",
    "from sklearn.base import BaseEstimator\n",
    "from sklearn.base import TransformerMixin\n",
    "\n",
    "# from sklearn.neural_network import MLPClassifier\n",
    "from sklearn.pipeline import make_pipeline\n",
    "\n",
    "# from xgboost import XGBClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "\n",
    "class ROIsFeatureExtractor(BaseEstimator, TransformerMixin):\n",
    "    \"\"\"Select only the 284 ROIs features:\"\"\"\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        return self\n",
    "\n",
    "    def transform(self, X):\n",
    "        return X[:, :284]\n",
    "\n",
    "\n",
    "def get_estimator():\n",
    "    \"\"\"Build your estimator here.\"\"\"\n",
    "    estimator = make_pipeline(\n",
    "        ROIsFeatureExtractor(), RandomForestClassifier(n_estimators=100, max_depth=5)\n",
    "    )\n",
    "\n",
    "    return estimator\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run locally:\n",
    "\n",
    "```\n",
    "ramp-test --submission starting_kit\n",
    "```\n",
    "\n",
    "## Submission on RAMP:\n",
    "\n",
    "Follow the instructions: [Submitting to a ramp event](https://paris-saclay-cds.github.io/ramp-docs/ramp-workflow/stable/using_kits.html#submitting-to-a-ramp-event)\n",
    "\n",
    "The event is [Predict schizophrenia from brain grey matter (classification)](https://ramp.studio/problems/brain_anatomy_schizophrenia)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ramp-test --submission starting_kit\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "python(41988) MallocStackLogging: can't turn off malloc stack logging because it was not enabled.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[38;5;178m\u001b[1mTesting Predict schizophrenia from     brain grey matter (classification)\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mReading train and test files from ./data/ ...\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mReading cv ...\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mTraining submissions/xgb_kenza-ily ...\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mCV fold 0\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc      time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m  \u001b[38;5;10m\u001b[1m1.00\u001b[0m  \u001b[38;5;150m0.98\u001b[0m  \u001b[38;5;150m0.875804\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.88\u001b[0m  \u001b[38;5;105m0.83\u001b[0m  \u001b[38;5;105m0.002712\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.83\u001b[0m  \u001b[38;5;218m0.79\u001b[0m  \u001b[38;5;218m0.001804\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mCV fold 1\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc      time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m  \u001b[38;5;10m\u001b[1m1.00\u001b[0m  \u001b[38;5;150m0.97\u001b[0m  \u001b[38;5;150m0.567521\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.71\u001b[0m  \u001b[38;5;105m0.61\u001b[0m  \u001b[38;5;105m0.002638\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.79\u001b[0m  \u001b[38;5;218m0.74\u001b[0m  \u001b[38;5;218m0.001630\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mCV fold 2\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc      time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m  \u001b[38;5;10m\u001b[1m1.00\u001b[0m  \u001b[38;5;150m0.98\u001b[0m  \u001b[38;5;150m0.558834\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.80\u001b[0m  \u001b[38;5;105m0.76\u001b[0m  \u001b[38;5;105m0.002621\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.78\u001b[0m  \u001b[38;5;218m0.72\u001b[0m  \u001b[38;5;218m0.001624\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mCV fold 3\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc      time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m  \u001b[38;5;10m\u001b[1m1.00\u001b[0m  \u001b[38;5;150m0.97\u001b[0m  \u001b[38;5;150m0.557006\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.77\u001b[0m  \u001b[38;5;105m0.76\u001b[0m  \u001b[38;5;105m0.002621\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.79\u001b[0m  \u001b[38;5;218m0.72\u001b[0m  \u001b[38;5;218m0.001647\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mCV fold 4\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc      time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m  \u001b[38;5;10m\u001b[1m1.00\u001b[0m  \u001b[38;5;150m0.96\u001b[0m  \u001b[38;5;150m0.560580\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.79\u001b[0m  \u001b[38;5;105m0.73\u001b[0m  \u001b[38;5;105m0.002591\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.80\u001b[0m  \u001b[38;5;218m0.74\u001b[0m  \u001b[38;5;218m0.001713\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1m----------------------------\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mMean CV scores\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1m----------------------------\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore           auc          bacc        time\u001b[0m\n",
      "\t\u001b[38;5;10m\u001b[1mtrain\u001b[0m   \u001b[38;5;10m\u001b[1m1.0\u001b[0m \u001b[38;5;150m\u001b[38;5;150m\u001b[38;5;150m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;150m0.001\u001b[0m  \u001b[38;5;150m0.97\u001b[0m \u001b[38;5;150m\u001b[38;5;150m\u001b[38;5;150m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;150m0.006\u001b[0m  \u001b[38;5;150m0.6\u001b[0m \u001b[38;5;150m\u001b[38;5;150m\u001b[38;5;150m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;150m0.13\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.79\u001b[0m \u001b[38;5;105m\u001b[38;5;105m\u001b[38;5;105m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;105m\u001b[38;5;105m\u001b[38;5;105m0.0\u001b[0m\u001b[0m55\u001b[0m  \u001b[38;5;105m0.74\u001b[0m \u001b[38;5;105m\u001b[38;5;105m\u001b[38;5;105m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;105m\u001b[38;5;105m\u001b[38;5;105m0.0\u001b[0m\u001b[0m71\u001b[0m   \u001b[38;5;105m\u001b[38;5;105m0.0\u001b[0m\u001b[0m \u001b[38;5;105m\u001b[38;5;105m\u001b[38;5;105m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;105m\u001b[38;5;105m0.0\u001b[0m\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m    \u001b[38;5;1m\u001b[1m0.8\u001b[0m \u001b[38;5;218m\u001b[38;5;218m\u001b[38;5;218m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;218m\u001b[38;5;218m\u001b[38;5;218m0.0\u001b[0m\u001b[0m17\u001b[0m  \u001b[38;5;218m0.74\u001b[0m \u001b[38;5;218m\u001b[38;5;218m\u001b[38;5;218m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;218m\u001b[38;5;218m\u001b[38;5;218m0.0\u001b[0m\u001b[0m25\u001b[0m   \u001b[38;5;218m\u001b[38;5;218m0.0\u001b[0m\u001b[0m \u001b[38;5;218m\u001b[38;5;218m\u001b[38;5;218m±\u001b[0m\u001b[0m\u001b[0m \u001b[38;5;218m\u001b[38;5;218m0.0\u001b[0m\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1m----------------------------\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1mBagged scores\u001b[0m\n",
      "\u001b[38;5;178m\u001b[1m----------------------------\u001b[0m\n",
      "\t\u001b[38;5;178m\u001b[1mscore   auc  bacc\u001b[0m\n",
      "\t\u001b[38;5;12m\u001b[1mvalid\u001b[0m  \u001b[38;5;12m\u001b[1m0.79\u001b[0m  \u001b[38;5;105m0.74\u001b[0m\n",
      "\t\u001b[38;5;1m\u001b[1mtest\u001b[0m   \u001b[38;5;1m\u001b[1m0.81\u001b[0m  \u001b[38;5;218m0.74\u001b[0m\n"
     ]
    }
   ],
   "source": [
    "!ramp-test --submission xgb_kenza-ily\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.2"
  },
  "vscode": {
   "interpreter": {
    "hash": "b69367434d80bac89d1a565c5caac4c5610f915c73dfb494ea62edf5b2129e8d"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}