Scripts/RandomForest_HumanSimulation.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Machine Learning applied to a simulated genetic dataset for a human disease\n",
    "For the Human dataset, I relied on simulations based on parameters I found in the literature.\n",
    "\n",
    "I simulated 4000 thousands cases and controls for 1M genetic variants, 100 of them simulating a disease.\n",
    "\n",
    "I then conducted a Genome Wide Association Study (GWAS), which identified most of the disease-related variants.\n",
    "\n",
    "Here, I'm describing the implementation of a `Random forest Classifier` on this dataset and then expand the prediction to a global dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import packages and data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib\n",
    "from matplotlib.colors import ListedColormap\n",
    "\n",
    "from pandas_plink import read_plink1_bin #package to read plink genetic variant files\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import metrics\n",
    "\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "sns.set()\n",
    "matplotlib.rcParams['figure.dpi'] = 200"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here I am loading the training set, which contains the top 25000 genetic variants as identified in the Genome Wide Association Study (GWAS).\n",
    "\n",
    "These variants are all SNPs, for Single Nucleotide Polymorphism, the most common type of variants.\n",
    "\n",
    "The reason I am not running the ML on the full 1 million SNPs is because of computation time and memory issues. Therefore, I decided to first run a GWAS and selected the top 25000 SNPs by P-value. Within these 25000 SNPs, only 100 are simulating a disease."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<xarray.DataArray 'genotype' (sample: 2800, variant: 25002)>\n",
      "dask.array<transpose, shape=(2800, 25002), dtype=float64, chunksize=(1024, 1024), chunktype=numpy.ndarray>\n",
      "Coordinates:\n",
      "  * sample   (sample) object 'per0' 'per1' 'per2' ... 'per3998' 'per3999'\n",
      "  * variant  (variant) object '1_neutral_D_191527' ... '22_neutral_D_47588'\n",
      "    fid      (sample) <U7 'per0' 'per1' 'per2' ... 'per3996' 'per3998' 'per3999'\n",
      "    iid      (sample) <U7 'per0' 'per1' 'per2' ... 'per3996' 'per3998' 'per3999'\n",
      "    father   (sample) <U1 '0' '0' '0' '0' '0' '0' ... '0' '0' '0' '0' '0' '0'\n",
      "    mother   (sample) <U1 '0' '0' '0' '0' '0' '0' ... '0' '0' '0' '0' '0' '0'\n",
      "    gender   (sample) <U1 '2' '2' '2' '2' '2' '2' ... '2' '2' '2' '2' '2' '2'\n",
      "    trait    (sample) float64 2.0 2.0 2.0 2.0 2.0 2.0 ... 1.0 1.0 1.0 1.0 1.0\n",
      "    chrom    (variant) <U2 '1' '1' '1' '1' '1' '1' ... '22' '22' '22' '22' '22'\n",
      "    snp      (variant) <U16 'neutral_D_191527' ... 'neutral_D_47588'\n",
      "    cm       (variant) float64 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0\n",
      "    pos      (variant) int64 1011209 1031563 1039813 ... 49205679 49411282\n",
      "    a0       (variant) <U1 'G' 'G' 'G' 'A' 'T' 'G' ... 'C' 'A' 'T' 'A' 'C' 'C'\n",
      "    a1       (variant) <U1 'A' 'A' 'A' 'G' 'G' 'A' ... 'A' 'G' 'C' 'G' 'G' 'T'\n"
     ]
    }
   ],
   "source": [
    "G_train = read_plink1_bin(\"Data/sim_ceu_train_top.bed\", \"Data/sim_ceu_train_top.bim\", \n",
    "                    \"Data/sim_ceu_train_top.fam\", verbose=False)\n",
    "print(G_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "case (2) /controls (1): [2. 2. 2. ... 1. 1. 1.]\n",
      "Size of training set (number of samples): 2800\n",
      "Number of controls: 1415\n",
      "Number of cases: 1385\n"
     ]
    }
   ],
   "source": [
    "y_train=np.array(G_train.trait)\n",
    "print('case (2) /controls (1): ' + str(y_train))\n",
    "print('Size of training set (number of samples):', len(y_train))\n",
    "print('Number of controls:', len(y_train[y_train==1]))\n",
    "print('Number of cases:', len(y_train[y_train==2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train=np.array(G_train.values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Size of testing set (number of samples): 1200\n",
      "Number of controls: 585\n",
      "Number of cases: 615\n"
     ]
    }
   ],
   "source": [
    "G_test = read_plink1_bin(\"Data/sim_ceu_test_top.bed\", \"Data/sim_ceu_test_top.bim\", \n",
    "                    \"Data/sim_ceu_test_top.fam\", verbose=False)\n",
    "y_test=np.array(G_test.trait)\n",
    "print('Size of testing set (number of samples):', len(y_test))\n",
    "print('Number of controls:', len(y_test[y_test==1]))\n",
    "print('Number of cases:', len(y_test[y_test==2]))\n",
    "X_test=np.array(G_test.values)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Results from the GWAS, ordered by P-value\n",
    "\n",
    "Important columns are \n",
    "* chr: chromosome\n",
    "* snp: name of SNP\n",
    "* bp: basepair, position in chr\n",
    "* P: P-value"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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>index</th>\n",
       "      <th>chr</th>\n",
       "      <th>snp</th>\n",
       "      <th>bp</th>\n",
       "      <th>A1</th>\n",
       "      <th>F_A</th>\n",
       "      <th>F_U</th>\n",
       "      <th>A2</th>\n",
       "      <th>CHISQ</th>\n",
       "      <th>P</th>\n",
       "      <th>OR</th>\n",
       "      <th>logp</th>\n",
       "      <th>chr_snp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>25001</td>\n",
       "      <td>6</td>\n",
       "      <td>disease_he_15</td>\n",
       "      <td>9952975</td>\n",
       "      <td>C</td>\n",
       "      <td>0.16750</td>\n",
       "      <td>0.08657</td>\n",
       "      <td>T</td>\n",
       "      <td>82.930</td>\n",
       "      <td>8.511000e-20</td>\n",
       "      <td>2.1230</td>\n",
       "      <td>19.070019</td>\n",
       "      <td>6_disease_he_15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>25000</td>\n",
       "      <td>16</td>\n",
       "      <td>disease_he_30</td>\n",
       "      <td>67967324</td>\n",
       "      <td>C</td>\n",
       "      <td>0.13900</td>\n",
       "      <td>0.07173</td>\n",
       "      <td>T</td>\n",
       "      <td>67.380</td>\n",
       "      <td>2.236000e-16</td>\n",
       "      <td>2.0890</td>\n",
       "      <td>15.650528</td>\n",
       "      <td>16_disease_he_30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>24999</td>\n",
       "      <td>1</td>\n",
       "      <td>disease_he_41</td>\n",
       "      <td>234783582</td>\n",
       "      <td>C</td>\n",
       "      <td>0.16930</td>\n",
       "      <td>0.09576</td>\n",
       "      <td>T</td>\n",
       "      <td>66.040</td>\n",
       "      <td>4.420000e-16</td>\n",
       "      <td>1.9250</td>\n",
       "      <td>15.354578</td>\n",
       "      <td>1_disease_he_41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>24998</td>\n",
       "      <td>16</td>\n",
       "      <td>disease_he_3</td>\n",
       "      <td>5594041</td>\n",
       "      <td>C</td>\n",
       "      <td>0.15780</td>\n",
       "      <td>0.08834</td>\n",
       "      <td>A</td>\n",
       "      <td>62.680</td>\n",
       "      <td>2.428000e-15</td>\n",
       "      <td>1.9330</td>\n",
       "      <td>14.614751</td>\n",
       "      <td>16_disease_he_3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>24997</td>\n",
       "      <td>4</td>\n",
       "      <td>disease_he_17</td>\n",
       "      <td>152830399</td>\n",
       "      <td>G</td>\n",
       "      <td>0.10140</td>\n",
       "      <td>0.04664</td>\n",
       "      <td>A</td>\n",
       "      <td>61.540</td>\n",
       "      <td>4.336000e-15</td>\n",
       "      <td>2.3080</td>\n",
       "      <td>14.362911</td>\n",
       "      <td>4_disease_he_17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24997</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>neutral_B_57276</td>\n",
       "      <td>59157063</td>\n",
       "      <td>G</td>\n",
       "      <td>0.07545</td>\n",
       "      <td>0.06219</td>\n",
       "      <td>T</td>\n",
       "      <td>3.845</td>\n",
       "      <td>4.991000e-02</td>\n",
       "      <td>1.2310</td>\n",
       "      <td>1.301812</td>\n",
       "      <td>3_neutral_B_57276</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24998</th>\n",
       "      <td>5</td>\n",
       "      <td>19</td>\n",
       "      <td>neutral_B_120957</td>\n",
       "      <td>52923460</td>\n",
       "      <td>T</td>\n",
       "      <td>0.07545</td>\n",
       "      <td>0.06219</td>\n",
       "      <td>C</td>\n",
       "      <td>3.845</td>\n",
       "      <td>4.991000e-02</td>\n",
       "      <td>1.2310</td>\n",
       "      <td>1.301812</td>\n",
       "      <td>19_neutral_B_120957</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24999</th>\n",
       "      <td>6</td>\n",
       "      <td>22</td>\n",
       "      <td>neutral_B_4912</td>\n",
       "      <td>15849026</td>\n",
       "      <td>A</td>\n",
       "      <td>0.04693</td>\n",
       "      <td>0.05866</td>\n",
       "      <td>G</td>\n",
       "      <td>3.845</td>\n",
       "      <td>4.991000e-02</td>\n",
       "      <td>0.7903</td>\n",
       "      <td>1.301812</td>\n",
       "      <td>22_neutral_B_4912</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25000</th>\n",
       "      <td>7</td>\n",
       "      <td>22</td>\n",
       "      <td>neutral_A_498329</td>\n",
       "      <td>31205871</td>\n",
       "      <td>G</td>\n",
       "      <td>0.04693</td>\n",
       "      <td>0.05866</td>\n",
       "      <td>A</td>\n",
       "      <td>3.845</td>\n",
       "      <td>4.991000e-02</td>\n",
       "      <td>0.7903</td>\n",
       "      <td>1.301812</td>\n",
       "      <td>22_neutral_A_498329</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25001</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>neutral_B_133767</td>\n",
       "      <td>40672301</td>\n",
       "      <td>G</td>\n",
       "      <td>0.04693</td>\n",
       "      <td>0.05866</td>\n",
       "      <td>A</td>\n",
       "      <td>3.845</td>\n",
       "      <td>4.991000e-02</td>\n",
       "      <td>0.7903</td>\n",
       "      <td>1.301812</td>\n",
       "      <td>3_neutral_B_133767</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>25002 rows × 13 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       index  chr               snp         bp A1      F_A      F_U A2  \\\n",
       "0      25001    6     disease_he_15    9952975  C  0.16750  0.08657  T   \n",
       "1      25000   16     disease_he_30   67967324  C  0.13900  0.07173  T   \n",
       "2      24999    1     disease_he_41  234783582  C  0.16930  0.09576  T   \n",
       "3      24998   16      disease_he_3    5594041  C  0.15780  0.08834  A   \n",
       "4      24997    4     disease_he_17  152830399  G  0.10140  0.04664  A   \n",
       "...      ...  ...               ...        ... ..      ...      ... ..   \n",
       "24997      1    3   neutral_B_57276   59157063  G  0.07545  0.06219  T   \n",
       "24998      5   19  neutral_B_120957   52923460  T  0.07545  0.06219  C   \n",
       "24999      6   22    neutral_B_4912   15849026  A  0.04693  0.05866  G   \n",
       "25000      7   22  neutral_A_498329   31205871  G  0.04693  0.05866  A   \n",
       "25001      0    3  neutral_B_133767   40672301  G  0.04693  0.05866  A   \n",
       "\n",
       "        CHISQ             P      OR       logp              chr_snp  \n",
       "0      82.930  8.511000e-20  2.1230  19.070019      6_disease_he_15  \n",
       "1      67.380  2.236000e-16  2.0890  15.650528     16_disease_he_30  \n",
       "2      66.040  4.420000e-16  1.9250  15.354578      1_disease_he_41  \n",
       "3      62.680  2.428000e-15  1.9330  14.614751      16_disease_he_3  \n",
       "4      61.540  4.336000e-15  2.3080  14.362911      4_disease_he_17  \n",
       "...       ...           ...     ...        ...                  ...  \n",
       "24997   3.845  4.991000e-02  1.2310   1.301812    3_neutral_B_57276  \n",
       "24998   3.845  4.991000e-02  1.2310   1.301812  19_neutral_B_120957  \n",
       "24999   3.845  4.991000e-02  0.7903   1.301812    22_neutral_B_4912  \n",
       "25000   3.845  4.991000e-02  0.7903   1.301812  22_neutral_A_498329  \n",
       "25001   3.845  4.991000e-02  0.7903   1.301812   3_neutral_B_133767  \n",
       "\n",
       "[25002 rows x 13 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gwas=pd.read_table('Data/topSNPs.txt',header=None)\n",
    "gwas.columns=['chr','snp','bp','A1','F_A','F_U','A2','CHISQ','P','OR','logp']\n",
    "gwas['chr_snp']=gwas['chr'].map(str) + \"_\" + gwas['snp']\n",
    "gwas=gwas.sort_values(by=['logp'], ascending=False)\n",
    "gwas=gwas.reset_index()\n",
    "gwas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining functions and the pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I am running the `Random Forest Classifier` in a `Pipeline` but I also wanted to control the number of SNPs that are considered in the model. \n",
    "\n",
    "Here I'm building a class object to perform this filtering. \n",
    "\n",
    "It takes as input:\n",
    "* the number of SNPs we want to keep\n",
    "* the results from the GWAS, in order to select SNPs by P-value\n",
    "* the full list of SNP names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['6_disease_he_15', '16_disease_he_30', '1_disease_he_41', '16_disease_he_3', '4_disease_he_17']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[[9231], [21350], [1871], [20924], [7190]]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#getting information on SNPs\n",
    "snp_names=list(np.array(G_train.variant))\n",
    "snp_keep=list(gwas.chr_snp[0:5])\n",
    "print(snp_keep)\n",
    "find = lambda searchList, elem: [[i for i, x in enumerate(searchList) if x == e] for e in elem]\n",
    "find(snp_names,snp_keep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import base\n",
    "\n",
    "class subsnp(base.BaseEstimator, base.TransformerMixin):\n",
    "    \n",
    "    def __init__(self, n_snps, snp_assoc, snp_names):\n",
    "        self.n_snps = n_snps\n",
    "        self.snp_assoc = snp_assoc\n",
    "        self.snp_names = snp_names\n",
    "    \n",
    "    def fit(self, X, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, X):\n",
    "        # Return an array with the selected SNPs as X\n",
    "        find = lambda searchList, elem: [[i for i, x in enumerate(searchList) if x == e] for e in elem]\n",
    "        snp_ids=list(self.snp_assoc.chr_snp[0:self.n_snps])\n",
    "        snp_idx=find(self.snp_names,snp_ids)\n",
    "        snp_idx=[y for x in snp_idx for y in x]\n",
    "        X=X[:,snp_idx]\n",
    "        return X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[1. 2. 2. ... 2. 2. 1.]\n",
      " [1. 2. 2. ... 2. 2. 2.]\n",
      " [2. 2. 1. ... 2. 2. 1.]\n",
      " ...\n",
      " [2. 1. 2. ... 2. 2. 1.]\n",
      " [2. 2. 2. ... 2. 2. 2.]\n",
      " [2. 0. 2. ... 2. 2. 2.]]\n"
     ]
    }
   ],
   "source": [
    "#testing the function\n",
    "snp_names=list(np.array(G_train.variant))\n",
    "snp_keep=list(gwas.chr_snp[0:5])\n",
    "n_snps=10\n",
    "subsnp_transform = subsnp(n_snps, gwas, snp_names)\n",
    "t=subsnp_transform.fit_transform(X_train)\n",
    "print(t)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['memory', 'steps', 'verbose', 'subsnp', 'reg', 'subsnp__n_snps', 'subsnp__snp_assoc', 'subsnp__snp_names', 'reg__bootstrap', 'reg__class_weight', 'reg__criterion', 'reg__max_depth', 'reg__max_features', 'reg__max_leaf_nodes', 'reg__min_impurity_decrease', 'reg__min_impurity_split', 'reg__min_samples_leaf', 'reg__min_samples_split', 'reg__min_weight_fraction_leaf', 'reg__n_estimators', 'reg__n_jobs', 'reg__oob_score', 'reg__random_state', 'reg__verbose', 'reg__warm_start'])"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "pipe_rfc = Pipeline([\n",
    "        # ColumnSelectTransformer\n",
    "    ('subsnp',subsnp(n_snps, gwas, snp_names)),\n",
    "        # KNeighborsRegressor\n",
    "    ('reg',RandomForestClassifier())\n",
    "    ])\n",
    "pipe_rfc.get_params().keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7385714285714285\n",
      "0.6208333333333333\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/data3/lib/python3.7/site-packages/sklearn/ensemble/forest.py:245: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "# testing the piepline\n",
    "pipe_rfc.fit(X_train, y_train)\n",
    "print(pipe_rfc.score(X_train, y_train))\n",
    "print(pipe_rfc.score(X_test, y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`Simple Imputer` would be useful to fix missing values, which is very common in genetic datasets.\n",
    "However, this is a simulation so it is not needed. Just putting it out here for future use."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.impute import SimpleImputer\n",
    "imp_mf = SimpleImputer(missing_values=np.nan, strategy='most_frequent')\n",
    "print(SimpleImputer())\n",
    "imp_mf.fit(X_train)\n",
    "X_train=imp_mf.transform(X_train)\n",
    "imp_mf.fit(X_test)\n",
    "X_test=imp_mf.transform(X_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Another transformation we may consider is `OneHotEncoding` the genetic data.\n",
    "\n",
    "As humans are diploid (i.e. 2x23 chromosomes), genetic variants can have values of 0, 1 or 2\n",
    "\n",
    "It did not boost performances in this project but I thought it might be useful for other datasets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# enc = OneHotEncoder(handle_unknown='ignore')\n",
    "# enc.fit(X_train)\n",
    "# X_train=enc.transform(X_train).toarray()\n",
    "# print(X_train.shape)\n",
    "# X_test=enc.transform(X_test).toarray()\n",
    "# print(X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# enc.get_feature_names()\n",
    "# will need to map this back to the SNP names when we look at feature importance and further analysis after fitting the model."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'subsnp__n_snps': [20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 1000, 2000, 3000, 4000, 5000], 'reg__n_estimators': [1000, 2000], 'reg__max_depth': [20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 1000, 2000, 3000, 4000, 5000], 'reg__min_samples_split': [10, 20, 30, 40], 'reg__min_samples_leaf': [1, 2, 4, 8]}\n",
      "Total number of parameter combinations:  7200\n"
     ]
    }
   ],
   "source": [
    "## Randomized search and random forest\n",
    "# Number of SNPs\n",
    "n_snps = [int(x) for x in np.linspace(20, 200, num = 10)]; n_snps.extend([int(x) for x in np.linspace(1000, 5000, num = 5)])\n",
    "# Number of trees in random forest\n",
    "n_estimators = [int(x) for x in np.linspace(start = 1000, stop = 2000, num = 2)]\n",
    "# Maximum number of levels in tree\n",
    "max_depth = [int(x) for x in np.linspace(20, 200, num = 10)]; \n",
    "max_depth.extend([int(x) for x in np.linspace(1000, 5000, num = 5)])\n",
    "# Minimum number of samples required to split a node\n",
    "min_samples_split = [int(x) for x in np.linspace(start = 10, stop = 40, num = 4)]\n",
    "# Minimum number of samples required at each leaf node\n",
    "min_samples_leaf = [1, 2, 4, 8]\n",
    "# Create the random grid\n",
    "random_grid = {'subsnp__n_snps': n_snps,\n",
    "               'reg__n_estimators': n_estimators,\n",
    "               'reg__max_depth': max_depth,\n",
    "               'reg__min_samples_split': min_samples_split,\n",
    "               'reg__min_samples_leaf': min_samples_leaf}\n",
    "print(random_grid)\n",
    "\n",
    "nb_par=1\n",
    "for key, value in random_grid.items():\n",
    "    nb_par*=len(value)\n",
    "print('Total number of parameter combinations: ',nb_par)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Randomized search and random forest\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import accuracy_score, recall_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 2000 candidates, totalling 10000 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=4)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=4)]: Done   5 tasks      | elapsed:   39.5s\n",
      "[Parallel(n_jobs=4)]: Done  10 tasks      | elapsed:   46.2s\n",
      "[Parallel(n_jobs=4)]: Done  17 tasks      | elapsed:  1.4min\n",
      "[Parallel(n_jobs=4)]: Done  24 tasks      | elapsed:  1.7min\n",
      "[Parallel(n_jobs=4)]: Done  33 tasks      | elapsed:  1.9min\n",
      "[Parallel(n_jobs=4)]: Done  42 tasks      | elapsed:  2.1min\n",
      "[Parallel(n_jobs=4)]: Done  53 tasks      | elapsed:  2.3min\n",
      "[Parallel(n_jobs=4)]: Done  64 tasks      | elapsed:  2.7min\n",
      "[Parallel(n_jobs=4)]: Done  77 tasks      | elapsed:  3.0min\n",
      "[Parallel(n_jobs=4)]: Done  90 tasks      | elapsed:  4.0min\n",
      "[Parallel(n_jobs=4)]: Done 105 tasks      | elapsed:  5.0min\n",
      "[Parallel(n_jobs=4)]: Done 120 tasks      | elapsed:  5.9min\n",
      "[Parallel(n_jobs=4)]: Done 137 tasks      | elapsed:  6.8min\n",
      "[Parallel(n_jobs=4)]: Done 154 tasks      | elapsed:  7.2min\n",
      "[Parallel(n_jobs=4)]: Done 173 tasks      | elapsed:  7.9min\n",
      "[Parallel(n_jobs=4)]: Done 192 tasks      | elapsed:  8.2min\n",
      "[Parallel(n_jobs=4)]: Done 213 tasks      | elapsed:  9.2min\n",
      "[Parallel(n_jobs=4)]: Done 234 tasks      | elapsed:  9.8min\n",
      "[Parallel(n_jobs=4)]: Done 257 tasks      | elapsed: 10.9min\n",
      "[Parallel(n_jobs=4)]: Done 280 tasks      | elapsed: 11.4min\n",
      "[Parallel(n_jobs=4)]: Done 305 tasks      | elapsed: 13.2min\n",
      "[Parallel(n_jobs=4)]: Done 330 tasks      | elapsed: 14.1min\n",
      "[Parallel(n_jobs=4)]: Done 357 tasks      | elapsed: 15.5min\n",
      "[Parallel(n_jobs=4)]: Done 384 tasks      | elapsed: 17.1min\n",
      "[Parallel(n_jobs=4)]: Done 413 tasks      | elapsed: 18.5min\n",
      "[Parallel(n_jobs=4)]: Done 442 tasks      | elapsed: 20.2min\n",
      "[Parallel(n_jobs=4)]: Done 473 tasks      | elapsed: 22.3min\n",
      "[Parallel(n_jobs=4)]: Done 504 tasks      | elapsed: 24.4min\n",
      "[Parallel(n_jobs=4)]: Done 537 tasks      | elapsed: 25.6min\n",
      "[Parallel(n_jobs=4)]: Done 570 tasks      | elapsed: 26.6min\n",
      "[Parallel(n_jobs=4)]: Done 605 tasks      | elapsed: 28.9min\n",
      "[Parallel(n_jobs=4)]: Done 640 tasks      | elapsed: 29.6min\n",
      "[Parallel(n_jobs=4)]: Done 677 tasks      | elapsed: 30.8min\n",
      "[Parallel(n_jobs=4)]: Done 714 tasks      | elapsed: 32.8min\n",
      "[Parallel(n_jobs=4)]: Done 753 tasks      | elapsed: 36.8min\n",
      "[Parallel(n_jobs=4)]: Done 792 tasks      | elapsed: 38.8min\n",
      "[Parallel(n_jobs=4)]: Done 833 tasks      | elapsed: 41.1min\n",
      "[Parallel(n_jobs=4)]: Done 874 tasks      | elapsed: 44.2min\n",
      "[Parallel(n_jobs=4)]: Done 917 tasks      | elapsed: 46.0min\n",
      "[Parallel(n_jobs=4)]: Done 960 tasks      | elapsed: 48.9min\n",
      "[Parallel(n_jobs=4)]: Done 1005 tasks      | elapsed: 50.4min\n",
      "[Parallel(n_jobs=4)]: Done 1050 tasks      | elapsed: 51.6min\n",
      "[Parallel(n_jobs=4)]: Done 1097 tasks      | elapsed: 53.5min\n",
      "[Parallel(n_jobs=4)]: Done 1144 tasks      | elapsed: 55.7min\n",
      "[Parallel(n_jobs=4)]: Done 1193 tasks      | elapsed: 57.2min\n",
      "[Parallel(n_jobs=4)]: Done 1242 tasks      | elapsed: 59.5min\n",
      "[Parallel(n_jobs=4)]: Done 1293 tasks      | elapsed: 62.0min\n",
      "[Parallel(n_jobs=4)]: Done 1344 tasks      | elapsed: 63.6min\n",
      "[Parallel(n_jobs=4)]: Done 1397 tasks      | elapsed: 65.8min\n",
      "[Parallel(n_jobs=4)]: Done 1450 tasks      | elapsed: 67.2min\n",
      "[Parallel(n_jobs=4)]: Done 1505 tasks      | elapsed: 69.6min\n",
      "[Parallel(n_jobs=4)]: Done 1560 tasks      | elapsed: 71.6min\n",
      "[Parallel(n_jobs=4)]: Done 1617 tasks      | elapsed: 74.8min\n",
      "[Parallel(n_jobs=4)]: Done 1674 tasks      | elapsed: 77.0min\n",
      "[Parallel(n_jobs=4)]: Done 1733 tasks      | elapsed: 79.2min\n",
      "[Parallel(n_jobs=4)]: Done 1792 tasks      | elapsed: 82.3min\n",
      "[Parallel(n_jobs=4)]: Done 1853 tasks      | elapsed: 84.3min\n",
      "[Parallel(n_jobs=4)]: Done 1914 tasks      | elapsed: 87.6min\n",
      "[Parallel(n_jobs=4)]: Done 1977 tasks      | elapsed: 90.5min\n",
      "[Parallel(n_jobs=4)]: Done 2040 tasks      | elapsed: 92.5min\n",
      "[Parallel(n_jobs=4)]: Done 2105 tasks      | elapsed: 95.6min\n",
      "[Parallel(n_jobs=4)]: Done 2170 tasks      | elapsed: 97.7min\n",
      "[Parallel(n_jobs=4)]: Done 2237 tasks      | elapsed: 100.1min\n",
      "[Parallel(n_jobs=4)]: Done 2304 tasks      | elapsed: 103.0min\n",
      "[Parallel(n_jobs=4)]: Done 2373 tasks      | elapsed: 106.7min\n",
      "[Parallel(n_jobs=4)]: Done 2442 tasks      | elapsed: 110.3min\n",
      "[Parallel(n_jobs=4)]: Done 2513 tasks      | elapsed: 114.3min\n",
      "[Parallel(n_jobs=4)]: Done 2584 tasks      | elapsed: 119.1min\n",
      "[Parallel(n_jobs=4)]: Done 2657 tasks      | elapsed: 122.4min\n",
      "[Parallel(n_jobs=4)]: Done 2730 tasks      | elapsed: 125.4min\n",
      "[Parallel(n_jobs=4)]: Done 2805 tasks      | elapsed: 127.6min\n",
      "[Parallel(n_jobs=4)]: Done 2880 tasks      | elapsed: 131.0min\n",
      "[Parallel(n_jobs=4)]: Done 2957 tasks      | elapsed: 133.5min\n",
      "[Parallel(n_jobs=4)]: Done 3034 tasks      | elapsed: 138.0min\n",
      "[Parallel(n_jobs=4)]: Done 3113 tasks      | elapsed: 142.3min\n",
      "[Parallel(n_jobs=4)]: Done 3192 tasks      | elapsed: 146.6min\n",
      "[Parallel(n_jobs=4)]: Done 3273 tasks      | elapsed: 149.9min\n",
      "[Parallel(n_jobs=4)]: Done 3354 tasks      | elapsed: 153.3min\n",
      "[Parallel(n_jobs=4)]: Done 3437 tasks      | elapsed: 157.8min\n",
      "[Parallel(n_jobs=4)]: Done 3520 tasks      | elapsed: 163.9min\n",
      "[Parallel(n_jobs=4)]: Done 3605 tasks      | elapsed: 167.7min\n",
      "[Parallel(n_jobs=4)]: Done 3690 tasks      | elapsed: 172.7min\n",
      "[Parallel(n_jobs=4)]: Done 3777 tasks      | elapsed: 177.1min\n",
      "[Parallel(n_jobs=4)]: Done 3864 tasks      | elapsed: 181.2min\n",
      "[Parallel(n_jobs=4)]: Done 3953 tasks      | elapsed: 185.7min\n",
      "[Parallel(n_jobs=4)]: Done 4042 tasks      | elapsed: 188.2min\n",
      "[Parallel(n_jobs=4)]: Done 4133 tasks      | elapsed: 194.5min\n",
      "[Parallel(n_jobs=4)]: Done 4224 tasks      | elapsed: 197.0min\n",
      "[Parallel(n_jobs=4)]: Done 4317 tasks      | elapsed: 200.5min\n",
      "[Parallel(n_jobs=4)]: Done 4410 tasks      | elapsed: 205.2min\n",
      "[Parallel(n_jobs=4)]: Done 4505 tasks      | elapsed: 208.8min\n",
      "[Parallel(n_jobs=4)]: Done 4600 tasks      | elapsed: 216.3min\n",
      "[Parallel(n_jobs=4)]: Done 4697 tasks      | elapsed: 220.7min\n",
      "[Parallel(n_jobs=4)]: Done 4794 tasks      | elapsed: 224.5min\n",
      "[Parallel(n_jobs=4)]: Done 4893 tasks      | elapsed: 229.5min\n",
      "[Parallel(n_jobs=4)]: Done 4992 tasks      | elapsed: 235.6min\n",
      "[Parallel(n_jobs=4)]: Done 5093 tasks      | elapsed: 240.6min\n",
      "[Parallel(n_jobs=4)]: Done 5194 tasks      | elapsed: 245.0min\n",
      "[Parallel(n_jobs=4)]: Done 5297 tasks      | elapsed: 252.0min\n",
      "[Parallel(n_jobs=4)]: Done 5400 tasks      | elapsed: 258.0min\n",
      "[Parallel(n_jobs=4)]: Done 5505 tasks      | elapsed: 264.2min\n",
      "[Parallel(n_jobs=4)]: Done 5610 tasks      | elapsed: 270.4min\n",
      "[Parallel(n_jobs=4)]: Done 5717 tasks      | elapsed: 273.9min\n",
      "[Parallel(n_jobs=4)]: Done 5824 tasks      | elapsed: 281.1min\n",
      "[Parallel(n_jobs=4)]: Done 5933 tasks      | elapsed: 286.9min\n",
      "[Parallel(n_jobs=4)]: Done 6042 tasks      | elapsed: 291.6min\n",
      "[Parallel(n_jobs=4)]: Done 6153 tasks      | elapsed: 298.9min\n",
      "/opt/conda/envs/data3/lib/python3.7/site-packages/joblib/externals/loky/process_executor.py:706: UserWarning: A worker stopped while some jobs were given to the executor. This can be caused by a too short worker timeout or by a memory leak.\n",
      "  \"timeout or by a memory leak.\", UserWarning\n",
      "[Parallel(n_jobs=4)]: Done 6264 tasks      | elapsed: 304.4min\n",
      "[Parallel(n_jobs=4)]: Done 6377 tasks      | elapsed: 309.2min\n",
      "[Parallel(n_jobs=4)]: Done 6490 tasks      | elapsed: 313.5min\n",
      "[Parallel(n_jobs=4)]: Done 6605 tasks      | elapsed: 318.8min\n",
      "[Parallel(n_jobs=4)]: Done 6720 tasks      | elapsed: 324.1min\n",
      "[Parallel(n_jobs=4)]: Done 6837 tasks      | elapsed: 329.1min\n",
      "[Parallel(n_jobs=4)]: Done 6954 tasks      | elapsed: 334.7min\n",
      "[Parallel(n_jobs=4)]: Done 7073 tasks      | elapsed: 340.1min\n",
      "[Parallel(n_jobs=4)]: Done 7192 tasks      | elapsed: 346.5min\n",
      "[Parallel(n_jobs=4)]: Done 7313 tasks      | elapsed: 353.6min\n",
      "[Parallel(n_jobs=4)]: Done 7434 tasks      | elapsed: 360.1min\n",
      "[Parallel(n_jobs=4)]: Done 7557 tasks      | elapsed: 363.3min\n",
      "[Parallel(n_jobs=4)]: Done 7680 tasks      | elapsed: 368.9min\n",
      "[Parallel(n_jobs=4)]: Done 7805 tasks      | elapsed: 375.4min\n",
      "[Parallel(n_jobs=4)]: Done 7930 tasks      | elapsed: 379.9min\n",
      "[Parallel(n_jobs=4)]: Done 8057 tasks      | elapsed: 385.5min\n",
      "[Parallel(n_jobs=4)]: Done 8184 tasks      | elapsed: 392.1min\n",
      "[Parallel(n_jobs=4)]: Done 8313 tasks      | elapsed: 399.4min\n",
      "[Parallel(n_jobs=4)]: Done 8442 tasks      | elapsed: 405.7min\n",
      "[Parallel(n_jobs=4)]: Done 8573 tasks      | elapsed: 410.7min\n",
      "[Parallel(n_jobs=4)]: Done 8704 tasks      | elapsed: 416.8min\n",
      "[Parallel(n_jobs=4)]: Done 8837 tasks      | elapsed: 421.7min\n",
      "[Parallel(n_jobs=4)]: Done 8970 tasks      | elapsed: 426.6min\n",
      "[Parallel(n_jobs=4)]: Done 9105 tasks      | elapsed: 433.7min\n",
      "[Parallel(n_jobs=4)]: Done 9240 tasks      | elapsed: 439.3min\n",
      "[Parallel(n_jobs=4)]: Done 9377 tasks      | elapsed: 446.1min\n",
      "[Parallel(n_jobs=4)]: Done 9514 tasks      | elapsed: 453.5min\n",
      "[Parallel(n_jobs=4)]: Done 9653 tasks      | elapsed: 460.5min\n",
      "[Parallel(n_jobs=4)]: Done 9792 tasks      | elapsed: 467.3min\n",
      "[Parallel(n_jobs=4)]: Done 9933 tasks      | elapsed: 474.3min\n",
      "[Parallel(n_jobs=4)]: Done 10000 out of 10000 | elapsed: 476.8min finished\n"
     ]
    }
   ],
   "source": [
    "rfc_random = RandomizedSearchCV(estimator = pipe_rfc, \n",
    "                               param_distributions = random_grid, \n",
    "                               n_iter = 2000, \n",
    "                               cv = 5, \n",
    "                               n_jobs = 4, \n",
    "                               verbose=10)\n",
    "\n",
    "# Fit the random search model\n",
    "rfc_random.fit(X_train, y_train)\n",
    "best_random = rfc_random.best_estimator_"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'subsnp__n_snps': 5000, 'reg__n_estimators': 2000, 'reg__min_samples_split': 20, 'reg__min_samples_leaf': 2, 'reg__max_depth': 2000}\n",
      "Accuracy: 0.74 Recall: 0.83\n"
     ]
    }
   ],
   "source": [
    "#Predict the response for test dataset\n",
    "print(rfc_random.best_params_)\n",
    "y_pred = best_random.predict(X_test)\n",
    "accuracy_rfc=metrics.accuracy_score(y_test, y_pred)\n",
    "recall_rfc=recall_score(y_test, y_pred)\n",
    "print(\"Accuracy:\",round(accuracy_rfc,2),\"Recall:\",round(recall_rfc,2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[487  98]\n",
      " [220 395]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         1.0       0.69      0.83      0.75       585\n",
      "         2.0       0.80      0.64      0.71       615\n",
      "\n",
      "    accuracy                           0.73      1200\n",
      "   macro avg       0.75      0.74      0.73      1200\n",
      "weighted avg       0.75      0.73      0.73      1200\n",
      "\n",
      "0.735\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
    "\n",
    "print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Accuracy: 1.0 Recall: 1.0\n",
      "Test Accuracy: 0.74 Recall: 0.83\n"
     ]
    }
   ],
   "source": [
    "#measure overfitting\n",
    "y_train_pred = best_random.predict(X_train)\n",
    "accuracy_train=metrics.accuracy_score(y_train, y_train_pred)\n",
    "recall_train=recall_score(y_train, y_train_pred)\n",
    "print(\"Train Accuracy:\",round(accuracy_train,2),\"Recall:\",round(recall_train,2))\n",
    "\n",
    "y_test_pred = best_random.predict(X_test)\n",
    "accuracy_test=metrics.accuracy_score(y_test, y_test_pred)\n",
    "recall_test=recall_score(y_test, y_test_pred)\n",
    "print(\"Test Accuracy:\",round(accuracy_test,2),\"Recall:\",round(recall_test,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The `Random Forest Classifier` obtained a fairly good result but suffers from overfitting. \n",
    "I attempted to reduce overfitting by increasing the number of `cross-validations` (from 3 to 5), splitting the dataset in a 70/30 ration instead of 80/20 and increasing the number of iterations of `RandomizedCV`.\n",
    "\n",
    "The result displayed here is the outcome of these modifications, which have, unfortunately, not reduced overfitting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['best_random_pipeline.joblib']"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Saving model\n",
    "#https://scikit-learn.org/stable/modules/model_persistence.html\n",
    "import pickle\n",
    "s = pickle.dumps(best_random)\n",
    "from joblib import dump, load\n",
    "dump(best_random, 'best_random_pipeline.joblib') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "#loading model\n",
    "from joblib import dump, load\n",
    "best_random=load('best_random_pipeline.joblib') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Comparison with GWAS and Feature importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "# feature importance\n",
    "importances = best_random.named_steps[\"reg\"].feature_importances_\n",
    "#std = np.std([best_random.feature_importances_ for tree in best_random.estimators_],\n",
    "#             axis=0)\n",
    "indices = np.argsort(importances)[::-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0            6_disease_he_15\n",
       "1           16_disease_he_30\n",
       "2            1_disease_he_41\n",
       "3            16_disease_he_3\n",
       "4            4_disease_he_17\n",
       "                ...         \n",
       "24997      3_neutral_B_57276\n",
       "24998    19_neutral_B_120957\n",
       "24999      22_neutral_B_4912\n",
       "25000    22_neutral_A_498329\n",
       "25001     3_neutral_B_133767\n",
       "Name: chr_snp, Length: 25002, dtype: object"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gwas.chr_snp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 2,  0,  3,  1,  8, 10,  9,  4,  7,  5, 12,  6, 16, 15, 17, 13, 18,\n",
       "       14, 11, 19])"
      ]
     },
     "execution_count": 90,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "indices[0:20]"
   ]
  },
  {
   "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>id</th>\n",
       "      <th>type</th>\n",
       "      <th>cat</th>\n",
       "      <th>variant</th>\n",
       "      <th>snp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>disease</td>\n",
       "      <td>he</td>\n",
       "      <td>41</td>\n",
       "      <td>6_disease_he_15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>disease</td>\n",
       "      <td>he</td>\n",
       "      <td>15</td>\n",
       "      <td>16_disease_he_30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>16</td>\n",
       "      <td>disease</td>\n",
       "      <td>he</td>\n",
       "      <td>3</td>\n",
       "      <td>1_disease_he_41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>16</td>\n",
       "      <td>disease</td>\n",
       "      <td>he</td>\n",
       "      <td>30</td>\n",
       "      <td>16_disease_he_3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>22</td>\n",
       "      <td>disease</td>\n",
       "      <td>he</td>\n",
       "      <td>20</td>\n",
       "      <td>4_disease_he_17</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id     type cat variant               snp\n",
       "0   1  disease  he      41   6_disease_he_15\n",
       "1   6  disease  he      15  16_disease_he_30\n",
       "2  16  disease  he       3   1_disease_he_41\n",
       "3  16  disease  he      30   16_disease_he_3\n",
       "4  22  disease  he      20   4_disease_he_17"
      ]
     },
     "execution_count": 91,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "topX=np.array(gwas.chr_snp[indices[0:59]])\n",
    "topX=pd.DataFrame([tmp.split('_') for tmp in topX],columns=['id','type','cat','variant'])\n",
    "topX['snp']=gwas.chr_snp[indices[0:59]]\n",
    "topX[0:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Comparing with GWAS, looking at the top 59 SNPs (59 SNPs were considered significant in the GWAS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "type     cat\n",
       "disease  he     46\n",
       "         ho      5\n",
       "neutral  B       1\n",
       "         C       1\n",
       "         D       6\n",
       "Name: variant, dtype: int64"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "topX.groupby(['type','cat'])['variant'].count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature ranking:\n",
      "1. 1_disease_he_41 (0.003094)\n",
      "2. 6_disease_he_15 (0.003016)\n",
      "3. 16_disease_he_3 (0.002440)\n",
      "4. 16_disease_he_30 (0.002405)\n",
      "5. 22_disease_he_20 (0.002333)\n",
      "6. 3_disease_he_25 (0.002170)\n",
      "7. 5_disease_he_11 (0.002159)\n",
      "8. 4_disease_he_17 (0.002145)\n",
      "9. 3_disease_he_45 (0.001974)\n",
      "10. 1_disease_he_32 (0.001935)\n"
     ]
    }
   ],
   "source": [
    "print(\"Feature ranking:\")\n",
    "for f in range(0,10):\n",
    "    print(\"%d. %s (%f)\" % (f + 1, topX.snp[indices[f]], importances[indices[f]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1200x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot the feature importances of the forest\n",
    "nb_feat=40\n",
    "plt.figure()\n",
    "plt.title(\"Feature importances\")\n",
    "plt.bar(range(0,nb_feat), importances[indices[0:nb_feat]],\n",
    "       color=\"r\", align=\"center\")\n",
    "plt.xticks(rotation=90)\n",
    "plt.xticks(range(0,nb_feat), topX.snp[indices[0:nb_feat]])\n",
    "plt.tick_params(axis='x', which='major', labelsize=6) \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "#roc curve, to do\n",
    "#https://towardsdatascience.com/an-implementation-and-explanation-of-the-random-forest-in-python-77bf308a9b76"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Global prediction \n",
    "\n",
    "Now that the model as been fitted on the simulated data, I wanted to predict the number of cases in an independent dataset.\n",
    "\n",
    "I used the genetic data from the [1000 genome project](https://www.internationalgenome.org/) in order to expand this prediction worldwide.\n",
    "\n",
    "The dataset I used contains 629 individuals from 12 different populations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<xarray.DataArray 'genotype' (sample: 629, variant: 25002)>\n",
      "dask.array<transpose, shape=(629, 25002), dtype=float64, chunksize=(629, 1024), chunktype=numpy.ndarray>\n",
      "Coordinates:\n",
      "  * sample   (sample) object 'HG00098' 'HG00100' ... 'NA20826' 'NA20828'\n",
      "  * variant  (variant) object '1_rs10907177' '1_rs6604968' ... '22_rs2071421'\n",
      "    fid      (sample) <U7 'HG00098' 'HG00100' 'HG00106' ... 'NA20826' 'NA20828'\n",
      "    iid      (sample) <U7 'HG00098' 'HG00100' 'HG00106' ... 'NA20826' 'NA20828'\n",
      "    father   (sample) <U1 '0' '0' '0' '0' '0' '0' ... '0' '0' '0' '0' '0' '0'\n",
      "    mother   (sample) <U1 '0' '0' '0' '0' '0' '0' ... '0' '0' '0' '0' '0' '0'\n",
      "    gender   (sample) <U1 '0' '0' '0' '0' '0' '0' ... '0' '0' '0' '0' '0' '0'\n",
      "    trait    (sample) float64 -9.0 -9.0 -9.0 -9.0 -9.0 ... -9.0 -9.0 -9.0 -9.0\n",
      "    chrom    (variant) <U2 '1' '1' '1' '1' '1' '1' ... '22' '22' '22' '22' '22'\n",
      "    snp      (variant) <U10 'rs10907177' 'rs6604968' ... 'rs7410608' 'rs2071421'\n",
      "    cm       (variant) float64 0.0 0.0 0.0 0.0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0\n",
      "    pos      (variant) int64 1011209 1031563 1039813 ... 49205679 49411282\n",
      "    a0       (variant) <U1 'G' 'G' 'G' 'A' 'T' 'G' ... 'C' 'A' 'T' 'A' 'C' 'C'\n",
      "    a1       (variant) <U1 'A' 'A' 'A' 'G' 'G' 'A' ... 'A' 'G' 'C' 'G' 'G' 'T'\n"
     ]
    }
   ],
   "source": [
    "#Load 1000 genome dataset\n",
    "kgp = read_plink1_bin(\"Data/1kG_top.bed\", \"Data/1kG_top.bim\", \"Data/1kG_top.fam\", verbose=False)\n",
    "print(kgp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(629, 25002)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_kgp=np.array(kgp.values)\n",
    "X_kgp.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SimpleImputer(add_indicator=False, copy=True, fill_value=None,\n",
      "              missing_values=nan, strategy='mean', verbose=0)\n"
     ]
    }
   ],
   "source": [
    "# Imputing missing values as most frequent\n",
    "# This time we have to impute given that it is a real dataset\n",
    "from sklearn.impute import SimpleImputer\n",
    "imp_mf = SimpleImputer(missing_values=np.nan, strategy='most_frequent')\n",
    "imp_mf.fit(X_kgp)\n",
    "print(SimpleImputer())\n",
    "X_kgp=imp_mf.transform(X_kgp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prediction for the 1000 genome dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "kpg_pred = best_random.predict(X_kgp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2. 2. 2. 2. 1. 2. 2. 2. 1. 2. 2. 2. 2. 2. 2. 2. 2. 2. 1. 1. 2. 2. 2. 2.\n",
      " 2. 2. 2. 2. 2. 2.]\n",
      "629\n"
     ]
    }
   ],
   "source": [
    "print(kpg_pred[0:30])\n",
    "print(len(kpg_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Adding the output of the model to a dataframe with population information"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.0    463\n",
       "1.0    166\n",
       "Name: Prediction, dtype: int64"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "kpg_df=pd.DataFrame(kpg_pred,columns=['Prediction'])\n",
    "kpg_df['Prediction'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {},
   "outputs": [],
   "source": [
    "igrs=pd.read_table('Data/igsr_samples.tsv') \n",
    "igrs = igrs[['Sample name','Sex','Population code','Superpopulation code','Population name']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(629, 6)\n"
     ]
    }
   ],
   "source": [
    "#get name of individuals from genetic dataset and combine with metadata\n",
    "kgp_sample=pd.DataFrame(np.array(kgp.sample),columns=['Sample name'])\n",
    "kpg_all=pd.merge(kgp_sample,igrs, on=['Sample name'])\n",
    "kpg_all['Prediction']=kpg_pred\n",
    "print(kpg_all.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TSI    92\n",
       "CEU    90\n",
       "JPT    84\n",
       "YRI    78\n",
       "CHB    68\n",
       "LWK    67\n",
       "GBR    43\n",
       "FIN    36\n",
       "CHS    25\n",
       "ASW    24\n",
       "MXL    17\n",
       "PUR     5\n",
       "Name: Population code, dtype: int64"
      ]
     },
     "execution_count": 105,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "kpg_all['Population code'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Remove PUR population because it only has 5 samples\n",
    "kpg_all=kpg_all[kpg_all['Population code']!='PUR']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "kpg_groupcode=pd.DataFrame(kpg_all.groupby(['Population code','Prediction'], as_index=False).count())\n",
    "kpg_groupcode=kpg_groupcode[['Population code','Prediction','Sample name']]\n",
    "#getting pop names related to codes\n",
    "code_desc=pd.DataFrame({'Population code': ['TSI','CEU','JPT','YRI','CHB','LWK','GBR','FIN','CHS','ASW','MXL'],\n",
    "                        'Population': ['Toscani','CEPH','Japanese','Yoruba','Han Chinese','Luhya','British','Finnish',\n",
    "        'Southern Han Chinese','African Ancestry SW','Mexican Ancestry']})\n",
    "\n",
    "kpg_groupcode2=pd.merge(kpg_groupcode,code_desc, on=['Population code'],how='left')\n",
    "kpg_groupcode2.loc[kpg_groupcode2.Prediction == 1, 'Prediction'] = \"Unaffected\"\n",
    "kpg_groupcode2.loc[kpg_groupcode2.Prediction == 2, 'Prediction'] = \"Affected\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Displaying the number of cases (in orange) and unaffected individuals per population"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1200x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#https://seaborn.pydata.org/generated/seaborn.barplot.html\n",
    "import seaborn as sns\n",
    "sns.set(style=\"whitegrid\")\n",
    "ax = sns.barplot(y=\"Population\", x=\"Sample name\", hue=\"Prediction\", data=kpg_groupcode2)\n",
    "ax.set_xlabel('count');\n",
    "ax.set_ylabel('');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gradient Boosting Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As an alternative to the `Random Forest Classifier`, I tried a `Gradient Boosting Classifier` in hope it would reduce overfitting. It also provides a valuable comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['memory', 'steps', 'verbose', 'subsnp', 'reg', 'subsnp__n_snps', 'subsnp__snp_assoc', 'subsnp__snp_names', 'reg__criterion', 'reg__init', 'reg__learning_rate', 'reg__loss', 'reg__max_depth', 'reg__max_features', 'reg__max_leaf_nodes', 'reg__min_impurity_decrease', 'reg__min_impurity_split', 'reg__min_samples_leaf', 'reg__min_samples_split', 'reg__min_weight_fraction_leaf', 'reg__n_estimators', 'reg__n_iter_no_change', 'reg__presort', 'reg__random_state', 'reg__subsample', 'reg__tol', 'reg__validation_fraction', 'reg__verbose', 'reg__warm_start'])"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "pipe_gbc = Pipeline([\n",
    "        # ColumnSelectTransformer\n",
    "    ('subsnp',subsnp(n_snps, gwas, snp_names)),\n",
    "        # KNeighborsRegressor\n",
    "    ('reg',GradientBoostingClassifier())\n",
    "    ])\n",
    "pipe_gbc.get_params().keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'subsnp__n_snps': [20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000], 'reg__n_estimators': [100, 200, 300, 400, 500, 600, 700, 800, 900, 1000], 'reg__max_depth': [20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000], 'reg__learning_rate': array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1. , 1.1, 1.2, 1.3,\n",
      "       1.4, 1.5, 1.6, 1.7, 1.8, 1.9]), 'reg__subsample': array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9])}\n"
     ]
    }
   ],
   "source": [
    "# Create the random grid\n",
    "# Number of SNPs\n",
    "n_snps = [int(x) for x in np.linspace(20, 200, num = 10)]; n_snps.extend([int(x) for x in np.linspace(500, 5000, num = 10)])\n",
    "# Number of trees in random forest\n",
    "n_estimators = [int(x) for x in np.linspace(start = 100, stop = 1000, num = 10)]\n",
    "# Maximum number of levels in tree\n",
    "max_depth = [int(x) for x in np.linspace(20, 200, num = 10)]; max_depth.extend([int(x) for x in np.linspace(500, 5000, num = 10)])\n",
    "# Minimum number of samples required to split a node\n",
    "min_samples_split = [int(x) for x in np.linspace(start = 10, stop = 100, num = 10)]\n",
    "# Minimum number of samples required at each leaf node\n",
    "min_samples_leaf = [1, 2, 4]\n",
    "# Create the random grid\n",
    "random_grid_gbc = {'subsnp__n_snps': n_snps,\n",
    "               'reg__n_estimators': n_estimators,\n",
    "               'reg__max_depth': max_depth,\n",
    "               'reg__learning_rate': np.arange(0.1,2,0.1),\n",
    "               'reg__subsample': np.arange(0.1,1,0.1)}\n",
    "print(random_grid_gbc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 200 candidates, totalling 600 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=4)]: Using backend LokyBackend with 4 concurrent workers.\n",
      "[Parallel(n_jobs=4)]: Done   5 tasks      | elapsed:   17.7s\n",
      "[Parallel(n_jobs=4)]: Done  10 tasks      | elapsed:  6.3min\n",
      "[Parallel(n_jobs=4)]: Done  17 tasks      | elapsed: 10.6min\n",
      "[Parallel(n_jobs=4)]: Done  24 tasks      | elapsed: 12.0min\n",
      "[Parallel(n_jobs=4)]: Done  33 tasks      | elapsed: 16.8min\n",
      "[Parallel(n_jobs=4)]: Done  42 tasks      | elapsed: 22.1min\n",
      "[Parallel(n_jobs=4)]: Done  53 tasks      | elapsed: 24.1min\n",
      "[Parallel(n_jobs=4)]: Done  64 tasks      | elapsed: 34.8min\n",
      "[Parallel(n_jobs=4)]: Done  77 tasks      | elapsed: 37.6min\n",
      "[Parallel(n_jobs=4)]: Done  90 tasks      | elapsed: 53.7min\n",
      "[Parallel(n_jobs=4)]: Done 105 tasks      | elapsed: 66.4min\n",
      "[Parallel(n_jobs=4)]: Done 120 tasks      | elapsed: 74.7min\n",
      "[Parallel(n_jobs=4)]: Done 137 tasks      | elapsed: 78.2min\n",
      "[Parallel(n_jobs=4)]: Done 154 tasks      | elapsed: 99.5min\n",
      "[Parallel(n_jobs=4)]: Done 173 tasks      | elapsed: 112.6min\n",
      "[Parallel(n_jobs=4)]: Done 192 tasks      | elapsed: 116.9min\n",
      "[Parallel(n_jobs=4)]: Done 213 tasks      | elapsed: 123.6min\n",
      "[Parallel(n_jobs=4)]: Done 234 tasks      | elapsed: 130.4min\n",
      "[Parallel(n_jobs=4)]: Done 257 tasks      | elapsed: 136.6min\n",
      "[Parallel(n_jobs=4)]: Done 280 tasks      | elapsed: 150.6min\n",
      "[Parallel(n_jobs=4)]: Done 305 tasks      | elapsed: 164.1min\n",
      "[Parallel(n_jobs=4)]: Done 330 tasks      | elapsed: 177.2min\n",
      "[Parallel(n_jobs=4)]: Done 357 tasks      | elapsed: 210.2min\n",
      "[Parallel(n_jobs=4)]: Done 384 tasks      | elapsed: 235.0min\n",
      "[Parallel(n_jobs=4)]: Done 413 tasks      | elapsed: 265.8min\n",
      "[Parallel(n_jobs=4)]: Done 442 tasks      | elapsed: 283.8min\n",
      "[Parallel(n_jobs=4)]: Done 473 tasks      | elapsed: 305.6min\n",
      "[Parallel(n_jobs=4)]: Done 504 tasks      | elapsed: 315.9min\n",
      "[Parallel(n_jobs=4)]: Done 537 tasks      | elapsed: 345.1min\n",
      "[Parallel(n_jobs=4)]: Done 570 tasks      | elapsed: 365.6min\n",
      "[Parallel(n_jobs=4)]: Done 600 out of 600 | elapsed: 377.0min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'subsnp__n_snps': 2000, 'reg__subsample': 0.30000000000000004, 'reg__n_estimators': 900, 'reg__max_depth': 1000, 'reg__learning_rate': 0.1}\n",
      "0.6458333333333334\n",
      "Accuracy: 0.65 Recall: 0.68\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "gbc_random = RandomizedSearchCV(estimator = pipe_gbc, \n",
    "                               param_distributions = random_grid_gbc, \n",
    "                               n_iter = 200, \n",
    "                               cv = 3, \n",
    "                               verbose=10,  \n",
    "                               n_jobs = 4)\n",
    "\n",
    "# Fit the random search model\n",
    "gbc_random.fit(X_train, y_train)\n",
    "print(gbc_random.best_params_)\n",
    "best_gbc_random = gbc_random.best_estimator_\n",
    "#Predict the response for test dataset\n",
    "y_pred = best_gbc_random.predict(X_test)\n",
    "print(gbc_random.score(X_test, y_test))\n",
    "accuracy2=metrics.accuracy_score(y_test, y_pred)\n",
    "recall2=recall_score(y_test, y_pred)\n",
    "print(\"Accuracy:\",round(accuracy2,2),\"Recall:\",round(recall2,2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[395 190]\n",
      " [235 380]]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         1.0       0.63      0.68      0.65       585\n",
      "         2.0       0.67      0.62      0.64       615\n",
      "\n",
      "    accuracy                           0.65      1200\n",
      "   macro avg       0.65      0.65      0.65      1200\n",
      "weighted avg       0.65      0.65      0.65      1200\n",
      "\n",
      "0.6458333333333334\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
    "\n",
    "print(confusion_matrix(y_test,y_pred))\n",
    "print(classification_report(y_test,y_pred))\n",
    "print(accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Accuracy: 1.0 Recall: 1.0\n",
      "Test Accuracy: 0.65 Recall: 0.68\n"
     ]
    }
   ],
   "source": [
    "#measure overfitting\n",
    "y_train_pred = best_gbc_random.predict(X_train)\n",
    "accuracy_train=metrics.accuracy_score(y_train, y_train_pred)\n",
    "recall_train=recall_score(y_train, y_train_pred)\n",
    "print(\"Train Accuracy:\",round(accuracy_train,2),\"Recall:\",round(recall_train,2))\n",
    "\n",
    "y_test_pred = best_gbc_random.predict(X_test)\n",
    "accuracy_test=metrics.accuracy_score(y_test, y_test_pred)\n",
    "recall_test=recall_score(y_test, y_test_pred)\n",
    "print(\"Test Accuracy:\",round(accuracy_test,2),\"Recall:\",round(recall_test,2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "My `Gradient Boosting Classifier` model also suffer overfitting and performed worse than the `Random Forest Classifier`"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}