06_LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_PATH = '../noaa-runtime/data/'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "dst = pd.read_csv(DATA_PATH +\"dst_labels.csv\")\n",
    "dst.timedelta = pd.to_timedelta(dst.timedelta)\n",
    "dst.set_index([\"period\", \"timedelta\"], inplace=True)\n",
    "\n",
    "sunspots = pd.read_csv(DATA_PATH + \"sunspots.csv\")\n",
    "sunspots.timedelta = pd.to_timedelta(sunspots.timedelta)\n",
    "sunspots.set_index([\"period\", \"timedelta\"], inplace=True)\n",
    "\n",
    "solar_wind = pd.read_csv(DATA_PATH + \"solar_wind.csv\")\n",
    "solar_wind.timedelta = pd.to_timedelta(solar_wind.timedelta)\n",
    "solar_wind.set_index([\"period\", \"timedelta\"], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from numpy.random import seed\n",
    "from tensorflow.random import set_seed\n",
    "\n",
    "seed(2020)\n",
    "set_seed(2021)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "# subset of solar wind features to use for modeling\n",
    "SOLAR_WIND_FEATURES = [ \"bt\", \"temperature\", \"bx_gse\", \"by_gse\",\n",
    "                       \"bz_gse\", \"speed\", \"density\"]\n",
    "\n",
    "# all of the features we'll use, including sunspot numbers\n",
    "XCOLS = ( [col + \"_mean\" for col in SOLAR_WIND_FEATURES]\n",
    "         + [col + \"_std\" for col in SOLAR_WIND_FEATURES]\n",
    "         + [\"smoothed_ssn\"] )\n",
    "\n",
    "\n",
    "def impute_features(feature_df):\n",
    "    \"\"\"Imputes data using the following methods:\n",
    "    - `smoothed_ssn`: forward fill\n",
    "    - `solar_wind`: interpolation\n",
    "    \"\"\"\n",
    "    # forward fill sunspot data for the rest of the month\n",
    "    feature_df.smoothed_ssn = feature_df.smoothed_ssn.fillna(method=\"ffill\")\n",
    "    # interpolate between missing solar wind values\n",
    "    feature_df = feature_df.interpolate()\n",
    "    return feature_df\n",
    "\n",
    "\n",
    "def aggregate_hourly(feature_df, aggs=[\"mean\", \"std\"]):\n",
    "    \"\"\"Aggregates features to the floor of each hour using mean and standard deviation.\n",
    "    e.g. All values from \"11:00:00\" to \"11:59:00\" will be aggregated to \"11:00:00\".\n",
    "    \"\"\"\n",
    "    # group by the floor of each hour use timedelta index\n",
    "    agged = feature_df.groupby(\n",
    "        [\"period\", feature_df.index.get_level_values(1).floor(\"H\")]\n",
    "    ).agg(aggs)\n",
    "    # flatten hierachical column index\n",
    "    agged.columns = [\"_\".join(x) for x in agged.columns]\n",
    "    return agged\n",
    "\n",
    "\n",
    "def preprocess_features(solar_wind, sunspots, scaler=None, subset=None):\n",
    "    \"\"\"\n",
    "    Preprocessing steps:\n",
    "        - Subset the data\n",
    "        - Aggregate hourly\n",
    "        - Join solar wind and sunspot data\n",
    "        - Scale using standard scaler\n",
    "        - Impute missing values\n",
    "    \"\"\"\n",
    "    # select features we want to use\n",
    "    if subset:\n",
    "        solar_wind = solar_wind[subset]\n",
    "\n",
    "    # aggregate solar wind data and join with sunspots\n",
    "    hourly_features = aggregate_hourly(solar_wind).join(sunspots)\n",
    "\n",
    "    # subtract mean and divide by standard deviation\n",
    "    if scaler is None:\n",
    "        scaler = StandardScaler()\n",
    "        scaler.fit(hourly_features)\n",
    "\n",
    "    normalized = pd.DataFrame(\n",
    "        scaler.transform(hourly_features),\n",
    "        index=hourly_features.index,\n",
    "        columns=hourly_features.columns,\n",
    "    )\n",
    "\n",
    "    # impute missing values\n",
    "    imputed = impute_features(normalized)\n",
    "\n",
    "    # we want to return the scaler object as well to use later during prediction\n",
    "    return imputed, scaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(139872, 15)\n"
     ]
    },
    {
     "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>bt_mean</th>\n",
       "      <th>bt_std</th>\n",
       "      <th>temperature_mean</th>\n",
       "      <th>temperature_std</th>\n",
       "      <th>bx_gse_mean</th>\n",
       "      <th>bx_gse_std</th>\n",
       "      <th>by_gse_mean</th>\n",
       "      <th>by_gse_std</th>\n",
       "      <th>bz_gse_mean</th>\n",
       "      <th>bz_gse_std</th>\n",
       "      <th>speed_mean</th>\n",
       "      <th>speed_std</th>\n",
       "      <th>density_mean</th>\n",
       "      <th>density_std</th>\n",
       "      <th>smoothed_ssn</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>period</th>\n",
       "      <th>timedelta</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",
       "      <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 rowspan=\"5\" valign=\"top\">train_a</th>\n",
       "      <th>00:00:00</th>\n",
       "      <td>0.499705</td>\n",
       "      <td>2.443614</td>\n",
       "      <td>-0.375267</td>\n",
       "      <td>0.383941</td>\n",
       "      <td>-1.599207</td>\n",
       "      <td>-0.381502</td>\n",
       "      <td>0.419516</td>\n",
       "      <td>0.031658</td>\n",
       "      <td>0.300358</td>\n",
       "      <td>-0.651645</td>\n",
       "      <td>-0.738546</td>\n",
       "      <td>0.862524</td>\n",
       "      <td>-0.775827</td>\n",
       "      <td>-0.205724</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>01:00:00</th>\n",
       "      <td>0.547177</td>\n",
       "      <td>-0.224580</td>\n",
       "      <td>-0.479430</td>\n",
       "      <td>0.953178</td>\n",
       "      <td>-1.757995</td>\n",
       "      <td>-0.867747</td>\n",
       "      <td>0.179257</td>\n",
       "      <td>-0.272971</td>\n",
       "      <td>0.446103</td>\n",
       "      <td>-0.517913</td>\n",
       "      <td>-0.986904</td>\n",
       "      <td>0.995063</td>\n",
       "      <td>-0.861692</td>\n",
       "      <td>-0.058215</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>02:00:00</th>\n",
       "      <td>0.739905</td>\n",
       "      <td>-0.770240</td>\n",
       "      <td>-0.574831</td>\n",
       "      <td>-0.192518</td>\n",
       "      <td>-1.912116</td>\n",
       "      <td>-1.114317</td>\n",
       "      <td>0.183266</td>\n",
       "      <td>-0.822786</td>\n",
       "      <td>0.770174</td>\n",
       "      <td>-0.876490</td>\n",
       "      <td>-1.013548</td>\n",
       "      <td>0.554085</td>\n",
       "      <td>-0.846222</td>\n",
       "      <td>-0.220012</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>03:00:00</th>\n",
       "      <td>0.699098</td>\n",
       "      <td>-0.278783</td>\n",
       "      <td>-0.324709</td>\n",
       "      <td>0.325491</td>\n",
       "      <td>-1.809045</td>\n",
       "      <td>-0.783042</td>\n",
       "      <td>-0.378111</td>\n",
       "      <td>0.341156</td>\n",
       "      <td>0.621194</td>\n",
       "      <td>-0.290211</td>\n",
       "      <td>-0.826469</td>\n",
       "      <td>-0.211185</td>\n",
       "      <td>-0.404306</td>\n",
       "      <td>0.218373</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>04:00:00</th>\n",
       "      <td>0.223933</td>\n",
       "      <td>-0.225168</td>\n",
       "      <td>-0.313432</td>\n",
       "      <td>0.201600</td>\n",
       "      <td>-1.338802</td>\n",
       "      <td>-0.484910</td>\n",
       "      <td>0.072745</td>\n",
       "      <td>1.023019</td>\n",
       "      <td>0.467629</td>\n",
       "      <td>-0.478080</td>\n",
       "      <td>-0.601238</td>\n",
       "      <td>1.016033</td>\n",
       "      <td>-0.371487</td>\n",
       "      <td>0.097253</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    bt_mean    bt_std  temperature_mean  temperature_std  \\\n",
       "period  timedelta                                                          \n",
       "train_a 00:00:00   0.499705  2.443614         -0.375267         0.383941   \n",
       "        01:00:00   0.547177 -0.224580         -0.479430         0.953178   \n",
       "        02:00:00   0.739905 -0.770240         -0.574831        -0.192518   \n",
       "        03:00:00   0.699098 -0.278783         -0.324709         0.325491   \n",
       "        04:00:00   0.223933 -0.225168         -0.313432         0.201600   \n",
       "\n",
       "                   bx_gse_mean  bx_gse_std  by_gse_mean  by_gse_std  \\\n",
       "period  timedelta                                                     \n",
       "train_a 00:00:00     -1.599207   -0.381502     0.419516    0.031658   \n",
       "        01:00:00     -1.757995   -0.867747     0.179257   -0.272971   \n",
       "        02:00:00     -1.912116   -1.114317     0.183266   -0.822786   \n",
       "        03:00:00     -1.809045   -0.783042    -0.378111    0.341156   \n",
       "        04:00:00     -1.338802   -0.484910     0.072745    1.023019   \n",
       "\n",
       "                   bz_gse_mean  bz_gse_std  speed_mean  speed_std  \\\n",
       "period  timedelta                                                   \n",
       "train_a 00:00:00      0.300358   -0.651645   -0.738546   0.862524   \n",
       "        01:00:00      0.446103   -0.517913   -0.986904   0.995063   \n",
       "        02:00:00      0.770174   -0.876490   -1.013548   0.554085   \n",
       "        03:00:00      0.621194   -0.290211   -0.826469  -0.211185   \n",
       "        04:00:00      0.467629   -0.478080   -0.601238   1.016033   \n",
       "\n",
       "                   density_mean  density_std  smoothed_ssn  \n",
       "period  timedelta                                           \n",
       "train_a 00:00:00      -0.775827    -0.205724      0.139444  \n",
       "        01:00:00      -0.861692    -0.058215      0.139444  \n",
       "        02:00:00      -0.846222    -0.220012      0.139444  \n",
       "        03:00:00      -0.404306     0.218373      0.139444  \n",
       "        04:00:00      -0.371487     0.097253      0.139444  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "features, scaler = preprocess_features(solar_wind, sunspots, subset=SOLAR_WIND_FEATURES)\n",
    "print(features.shape)\n",
    "features.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# check to make sure missing values are filled\n",
    "assert (features.isna().sum() == 0).all()"
   ]
  },
  {
   "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></th>\n",
       "      <th>t0</th>\n",
       "      <th>t1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>period</th>\n",
       "      <th>timedelta</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">train_a</th>\n",
       "      <th>00:00:00</th>\n",
       "      <td>-7</td>\n",
       "      <td>-10.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>01:00:00</th>\n",
       "      <td>-10</td>\n",
       "      <td>-10.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>02:00:00</th>\n",
       "      <td>-10</td>\n",
       "      <td>-6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>03:00:00</th>\n",
       "      <td>-6</td>\n",
       "      <td>-2.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>04:00:00</th>\n",
       "      <td>-2</td>\n",
       "      <td>3.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   t0    t1\n",
       "period  timedelta          \n",
       "train_a 00:00:00   -7 -10.0\n",
       "        01:00:00  -10 -10.0\n",
       "        02:00:00  -10  -6.0\n",
       "        03:00:00   -6  -2.0\n",
       "        04:00:00   -2   3.0"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "YCOLS = [\"t0\", \"t1\"]\n",
    "\n",
    "def process_labels(dst):\n",
    "    y = dst.copy()\n",
    "    y[\"t1\"] = y.groupby(\"period\").dst.shift(-1)\n",
    "    y.columns = YCOLS\n",
    "    return y\n",
    "\n",
    "labels = process_labels(dst)\n",
    "labels.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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></th>\n",
       "      <th>t0</th>\n",
       "      <th>t1</th>\n",
       "      <th>bt_mean</th>\n",
       "      <th>bt_std</th>\n",
       "      <th>temperature_mean</th>\n",
       "      <th>temperature_std</th>\n",
       "      <th>bx_gse_mean</th>\n",
       "      <th>bx_gse_std</th>\n",
       "      <th>by_gse_mean</th>\n",
       "      <th>by_gse_std</th>\n",
       "      <th>bz_gse_mean</th>\n",
       "      <th>bz_gse_std</th>\n",
       "      <th>speed_mean</th>\n",
       "      <th>speed_std</th>\n",
       "      <th>density_mean</th>\n",
       "      <th>density_std</th>\n",
       "      <th>smoothed_ssn</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>period</th>\n",
       "      <th>timedelta</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",
       "      <th></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 rowspan=\"5\" valign=\"top\">train_a</th>\n",
       "      <th>00:00:00</th>\n",
       "      <td>-7</td>\n",
       "      <td>-10.0</td>\n",
       "      <td>0.499705</td>\n",
       "      <td>2.443614</td>\n",
       "      <td>-0.375267</td>\n",
       "      <td>0.383941</td>\n",
       "      <td>-1.599207</td>\n",
       "      <td>-0.381502</td>\n",
       "      <td>0.419516</td>\n",
       "      <td>0.031658</td>\n",
       "      <td>0.300358</td>\n",
       "      <td>-0.651645</td>\n",
       "      <td>-0.738546</td>\n",
       "      <td>0.862524</td>\n",
       "      <td>-0.775827</td>\n",
       "      <td>-0.205724</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>01:00:00</th>\n",
       "      <td>-10</td>\n",
       "      <td>-10.0</td>\n",
       "      <td>0.547177</td>\n",
       "      <td>-0.224580</td>\n",
       "      <td>-0.479430</td>\n",
       "      <td>0.953178</td>\n",
       "      <td>-1.757995</td>\n",
       "      <td>-0.867747</td>\n",
       "      <td>0.179257</td>\n",
       "      <td>-0.272971</td>\n",
       "      <td>0.446103</td>\n",
       "      <td>-0.517913</td>\n",
       "      <td>-0.986904</td>\n",
       "      <td>0.995063</td>\n",
       "      <td>-0.861692</td>\n",
       "      <td>-0.058215</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>02:00:00</th>\n",
       "      <td>-10</td>\n",
       "      <td>-6.0</td>\n",
       "      <td>0.739905</td>\n",
       "      <td>-0.770240</td>\n",
       "      <td>-0.574831</td>\n",
       "      <td>-0.192518</td>\n",
       "      <td>-1.912116</td>\n",
       "      <td>-1.114317</td>\n",
       "      <td>0.183266</td>\n",
       "      <td>-0.822786</td>\n",
       "      <td>0.770174</td>\n",
       "      <td>-0.876490</td>\n",
       "      <td>-1.013548</td>\n",
       "      <td>0.554085</td>\n",
       "      <td>-0.846222</td>\n",
       "      <td>-0.220012</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>03:00:00</th>\n",
       "      <td>-6</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>0.699098</td>\n",
       "      <td>-0.278783</td>\n",
       "      <td>-0.324709</td>\n",
       "      <td>0.325491</td>\n",
       "      <td>-1.809045</td>\n",
       "      <td>-0.783042</td>\n",
       "      <td>-0.378111</td>\n",
       "      <td>0.341156</td>\n",
       "      <td>0.621194</td>\n",
       "      <td>-0.290211</td>\n",
       "      <td>-0.826469</td>\n",
       "      <td>-0.211185</td>\n",
       "      <td>-0.404306</td>\n",
       "      <td>0.218373</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>04:00:00</th>\n",
       "      <td>-2</td>\n",
       "      <td>3.0</td>\n",
       "      <td>0.223933</td>\n",
       "      <td>-0.225168</td>\n",
       "      <td>-0.313432</td>\n",
       "      <td>0.201600</td>\n",
       "      <td>-1.338802</td>\n",
       "      <td>-0.484910</td>\n",
       "      <td>0.072745</td>\n",
       "      <td>1.023019</td>\n",
       "      <td>0.467629</td>\n",
       "      <td>-0.478080</td>\n",
       "      <td>-0.601238</td>\n",
       "      <td>1.016033</td>\n",
       "      <td>-0.371487</td>\n",
       "      <td>0.097253</td>\n",
       "      <td>0.139444</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   t0    t1   bt_mean    bt_std  temperature_mean  \\\n",
       "period  timedelta                                                   \n",
       "train_a 00:00:00   -7 -10.0  0.499705  2.443614         -0.375267   \n",
       "        01:00:00  -10 -10.0  0.547177 -0.224580         -0.479430   \n",
       "        02:00:00  -10  -6.0  0.739905 -0.770240         -0.574831   \n",
       "        03:00:00   -6  -2.0  0.699098 -0.278783         -0.324709   \n",
       "        04:00:00   -2   3.0  0.223933 -0.225168         -0.313432   \n",
       "\n",
       "                   temperature_std  bx_gse_mean  bx_gse_std  by_gse_mean  \\\n",
       "period  timedelta                                                          \n",
       "train_a 00:00:00          0.383941    -1.599207   -0.381502     0.419516   \n",
       "        01:00:00          0.953178    -1.757995   -0.867747     0.179257   \n",
       "        02:00:00         -0.192518    -1.912116   -1.114317     0.183266   \n",
       "        03:00:00          0.325491    -1.809045   -0.783042    -0.378111   \n",
       "        04:00:00          0.201600    -1.338802   -0.484910     0.072745   \n",
       "\n",
       "                   by_gse_std  bz_gse_mean  bz_gse_std  speed_mean  speed_std  \\\n",
       "period  timedelta                                                               \n",
       "train_a 00:00:00     0.031658     0.300358   -0.651645   -0.738546   0.862524   \n",
       "        01:00:00    -0.272971     0.446103   -0.517913   -0.986904   0.995063   \n",
       "        02:00:00    -0.822786     0.770174   -0.876490   -1.013548   0.554085   \n",
       "        03:00:00     0.341156     0.621194   -0.290211   -0.826469  -0.211185   \n",
       "        04:00:00     1.023019     0.467629   -0.478080   -0.601238   1.016033   \n",
       "\n",
       "                   density_mean  density_std  smoothed_ssn  \n",
       "period  timedelta                                           \n",
       "train_a 00:00:00      -0.775827    -0.205724      0.139444  \n",
       "        01:00:00      -0.861692    -0.058215      0.139444  \n",
       "        02:00:00      -0.846222    -0.220012      0.139444  \n",
       "        03:00:00      -0.404306     0.218373      0.139444  \n",
       "        04:00:00      -0.371487     0.097253      0.139444  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = labels.join(features)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_test_val(data, test_per_period, val_per_period):\n",
    "    \"\"\"Splits data across periods into train, test, and validation\"\"\"\n",
    "    # assign the last `test_per_period` rows from each period to test\n",
    "    test = data.groupby(\"period\").tail(test_per_period)\n",
    "    interim = data[~data.index.isin(test.index)]\n",
    "    # assign the last `val_per_period` from the remaining rows to validation\n",
    "    val = interim.groupby(\"period\").tail(val_per_period)\n",
    "    # the remaining rows are assigned to train\n",
    "    train = interim[~interim.index.isin(val.index)]\n",
    "    return train, test, val\n",
    "\n",
    "train, test, val = get_train_test_val(data, test_per_period=6_000, val_per_period=3_000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plots():\n",
    "    \n",
    "    ind = [0, 1, 2]\n",
    "    names = [\"train_a\", \"train_b\", \"train_c\"]\n",
    "    width = 0.75\n",
    "    train_cnts = [len(df) for _, df in train.groupby(\"period\")]\n",
    "    val_cnts = [len(df) for _, df in val.groupby(\"period\")]\n",
    "    test_cnts = [len(df) for _, df in test.groupby(\"period\")]\n",
    "\n",
    "    p1 = plt.barh(ind, train_cnts, width)\n",
    "    p2 = plt.barh(ind, val_cnts, width, left=train_cnts)\n",
    "    p3 = plt.barh(ind, test_cnts, width, left=np.add(val_cnts, train_cnts).tolist())\n",
    "\n",
    "    plt.yticks(ind, names)\n",
    "    plt.ylabel(\"Period\")\n",
    "    plt.xlabel(\"Hourly Timesteps\")\n",
    "    plt.title(\"Train/Validation/Test Splits over Periods\", fontsize=16)\n",
    "    plt.legend([\"Train\", \"Validation\", \"Test\"])\n",
    "    \n",
    "plots()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, LSTM, Dropout\n",
    "from tensorflow.keras import preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of train batches: 3523\n",
      "Number of val batches: 276\n"
     ]
    }
   ],
   "source": [
    "data_config = { \"timesteps\": 32,\n",
    "               \"batch_size\": 32}\n",
    "\n",
    "def timeseries_dataset_from_df(df, batch_size):\n",
    "    dataset = None\n",
    "    timesteps = data_config[\"timesteps\"]\n",
    "\n",
    "    # iterate through periods\n",
    "    for _, period_df in df.groupby(\"period\"):\n",
    "        # realign features and labels so that first sequence of 32 is aligned with the 33rd target\n",
    "        inputs = period_df[XCOLS][:-timesteps]\n",
    "        outputs = period_df[YCOLS][timesteps:]\n",
    "\n",
    "        period_ds = preprocessing.timeseries_dataset_from_array(\n",
    "            inputs,\n",
    "            outputs,\n",
    "            timesteps,\n",
    "            batch_size=batch_size,\n",
    "        )\n",
    "\n",
    "        if dataset is None:\n",
    "            dataset = period_ds\n",
    "        else:\n",
    "            dataset = dataset.concatenate(period_ds)\n",
    "\n",
    "    return dataset\n",
    "\n",
    "\n",
    "train_ds = timeseries_dataset_from_df(train, data_config[\"batch_size\"])\n",
    "val_ds = timeseries_dataset_from_df(val, data_config[\"batch_size\"])\n",
    "\n",
    "print(f\"Number of train batches: {len(train_ds)}\")\n",
    "print(f\"Number of val batches: {len(val_ds)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm (LSTM)                  (None, 64)                20480     \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 2)                 66        \n",
      "=================================================================\n",
      "Total params: 26,786\n",
      "Trainable params: 26,786\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# define our model\n",
    "model_config = {\"n_epochs\": 50, \"n_neurons\": 64, \"dropout\": 0.4, \"stateful\": False}\n",
    "\n",
    "def make_model1():\n",
    "    model = Sequential()\n",
    "    model.add(\n",
    "        LSTM(\n",
    "            model_config[\"n_neurons\"],\n",
    "            # usually set to (`batch_size`, `sequence_length`, `n_features`)\n",
    "            # setting the batch size to None allows for variable length batches\n",
    "            batch_input_shape=(None, data_config[\"timesteps\"], len(XCOLS)),\n",
    "            stateful=model_config[\"stateful\"],\n",
    "            dropout=model_config[\"dropout\"],\n",
    "        )\n",
    "    )\n",
    "\n",
    "    model.add(Dense(64))\n",
    "    model.add(Dense(32))\n",
    "    \n",
    "    model.add(Dense(len(YCOLS)))\n",
    "    model.compile( loss=\"mean_squared_error\", optimizer=\"adam\")\n",
    "    return model\n",
    "\n",
    "model = make_model1()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for name, values in history.history.items():\n",
    "    plt.plot(values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_config = {\"n_epochs\": 10, \"n_neurons\": 64, \"dropout\": 0.4, \"stateful\": False}\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print ('n_epochs=',model_config[\"n_epochs\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#for name, values in history.history.items():\n",
    "#    plt.plot(values)\n",
    "#loss=\n",
    "loss=history.history['loss']\n",
    "val_loss=history.history['val_loss']\n",
    "plt.plot(range(1, len(loss)+1), loss, label='Loss' )\n",
    "plt.plot(range(1, len(loss)+1), val_loss, label='Val Loss' )\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_3 (LSTM)                (None, 32)                6144      \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 2)                 66        \n",
      "=================================================================\n",
      "Total params: 9,378\n",
      "Trainable params: 9,378\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# define our model\n",
    "model_config = {\"n_epochs\": 50, \"n_neurons\": 32, \"dropout\": 0.4, \"stateful\": False}\n",
    "\n",
    "def make_model2():\n",
    "    model = Sequential()\n",
    "    model.add(\n",
    "        LSTM(\n",
    "            model_config[\"n_neurons\"],\n",
    "            # usually set to (`batch_size`, `sequence_length`, `n_features`)\n",
    "            # setting the batch size to None allows for variable length batches\n",
    "            batch_input_shape=(None, data_config[\"timesteps\"], len(XCOLS)),\n",
    "            stateful=model_config[\"stateful\"],\n",
    "            dropout=model_config[\"dropout\"],\n",
    "        )\n",
    "    )\n",
    "    model.add(Dense(32))\n",
    "    model.add(Dense(32))\n",
    "    model.add(Dense(32))\n",
    "    model.add(Dense(len(YCOLS)))\n",
    "    model.compile(\n",
    "        loss=\"mean_squared_error\",\n",
    "        optimizer=\"adam\",\n",
    "    )\n",
    "    return model\n",
    "\n",
    "model = make_model2()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 202.2973 - val_loss: 314.4594\n",
      "Epoch 2/50\n",
      "3523/3523 [==============================] - 49s 14ms/step - loss: 166.1333 - val_loss: 263.7596\n",
      "Epoch 3/50\n",
      "3523/3523 [==============================] - 44s 13ms/step - loss: 154.9315 - val_loss: 242.8779\n",
      "Epoch 4/50\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 150.2044 - val_loss: 231.8553\n",
      "Epoch 5/50\n",
      "3523/3523 [==============================] - 47s 13ms/step - loss: 145.6048 - val_loss: 224.0678\n",
      "Epoch 6/50\n",
      "3523/3523 [==============================] - 43s 12ms/step - loss: 140.9312 - val_loss: 219.3269\n",
      "Epoch 7/50\n",
      "3523/3523 [==============================] - 43s 12ms/step - loss: 136.9265 - val_loss: 217.6194\n",
      "Epoch 8/50\n",
      "3523/3523 [==============================] - 44s 12ms/step - loss: 134.2207 - val_loss: 220.2895\n",
      "Epoch 9/50\n",
      "3523/3523 [==============================] - 44s 13ms/step - loss: 133.7776 - val_loss: 219.7985\n",
      "Epoch 10/50\n",
      "3523/3523 [==============================] - 44s 13ms/step - loss: 128.2733 - val_loss: 224.0457\n",
      "Epoch 11/50\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 126.3500 - val_loss: 230.2963\n",
      "Epoch 12/50\n",
      "3523/3523 [==============================] - 47s 13ms/step - loss: 125.5717 - val_loss: 221.9114\n",
      "Epoch 13/50\n",
      "3523/3523 [==============================] - 56s 16ms/step - loss: 124.2558 - val_loss: 241.7894\n",
      "Epoch 14/50\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 122.4132 - val_loss: 235.8404\n",
      "Epoch 15/50\n",
      "3523/3523 [==============================] - 44s 13ms/step - loss: 120.9576 - val_loss: 219.4011\n",
      "Epoch 16/50\n",
      "3523/3523 [==============================] - 44s 12ms/step - loss: 120.6057 - val_loss: 226.8738\n",
      "Epoch 17/50\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 116.9768 - val_loss: 231.9377\n",
      "Epoch 18/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 118.3890 - val_loss: 247.0023\n",
      "Epoch 19/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 118.7132 - val_loss: 226.3943\n",
      "Epoch 20/50\n",
      "3523/3523 [==============================] - 49s 14ms/step - loss: 116.9785 - val_loss: 221.1367\n",
      "Epoch 21/50\n",
      "3523/3523 [==============================] - 50s 14ms/step - loss: 116.9413 - val_loss: 219.1031\n",
      "Epoch 22/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 116.8244 - val_loss: 224.7138\n",
      "Epoch 23/50\n",
      "3523/3523 [==============================] - 47s 13ms/step - loss: 115.6732 - val_loss: 234.5882\n",
      "Epoch 24/50\n",
      "3523/3523 [==============================] - 52s 15ms/step - loss: 115.9998 - val_loss: 234.5753\n",
      "Epoch 25/50\n",
      "3523/3523 [==============================] - 51s 15ms/step - loss: 114.7815 - val_loss: 222.8411\n",
      "Epoch 26/50\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 115.3693 - val_loss: 220.0969\n",
      "Epoch 27/50\n",
      "3523/3523 [==============================] - 53s 15ms/step - loss: 113.8815 - val_loss: 215.1097\n",
      "Epoch 28/50\n",
      "3523/3523 [==============================] - 49s 14ms/step - loss: 113.2334 - val_loss: 212.6205\n",
      "Epoch 29/50\n",
      "3523/3523 [==============================] - 52s 15ms/step - loss: 113.6453 - val_loss: 212.6164\n",
      "Epoch 30/50\n",
      "3523/3523 [==============================] - 54s 15ms/step - loss: 112.2040 - val_loss: 209.6959\n",
      "Epoch 31/50\n",
      "3523/3523 [==============================] - 59s 17ms/step - loss: 111.4948 - val_loss: 209.3847\n",
      "Epoch 32/50\n",
      "3523/3523 [==============================] - 55s 16ms/step - loss: 110.5754 - val_loss: 205.8010\n",
      "Epoch 33/50\n",
      "3523/3523 [==============================] - 54s 15ms/step - loss: 108.4593 - val_loss: 201.5607\n",
      "Epoch 34/50\n",
      "3523/3523 [==============================] - 54s 15ms/step - loss: 110.1153 - val_loss: 223.6039\n",
      "Epoch 35/50\n",
      "3523/3523 [==============================] - 59s 17ms/step - loss: 108.9255 - val_loss: 220.4974\n",
      "Epoch 36/50\n",
      "3523/3523 [==============================] - 55s 16ms/step - loss: 108.1689 - val_loss: 210.5815\n",
      "Epoch 37/50\n",
      "3523/3523 [==============================] - 51s 14ms/step - loss: 108.7984 - val_loss: 224.4097\n",
      "Epoch 38/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 107.7572 - val_loss: 212.6693\n",
      "Epoch 39/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 107.2288 - val_loss: 211.7143\n",
      "Epoch 40/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 105.8851 - val_loss: 214.8994\n",
      "Epoch 41/50\n",
      "3523/3523 [==============================] - 47s 13ms/step - loss: 107.1641 - val_loss: 223.0328\n",
      "Epoch 42/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 106.1543 - val_loss: 232.8680\n",
      "Epoch 43/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 105.3223 - val_loss: 215.8983\n",
      "Epoch 44/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 105.3371 - val_loss: 221.9686\n",
      "Epoch 45/50\n",
      "3523/3523 [==============================] - 51s 14ms/step - loss: 105.1461 - val_loss: 213.2332\n",
      "Epoch 46/50\n",
      "3523/3523 [==============================] - 47s 13ms/step - loss: 105.6584 - val_loss: 216.0652\n",
      "Epoch 47/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 105.7539 - val_loss: 216.6619\n",
      "Epoch 48/50\n",
      "3523/3523 [==============================] - 42s 12ms/step - loss: 103.5330 - val_loss: 223.2279\n",
      "Epoch 49/50\n",
      "3523/3523 [==============================] - 44s 13ms/step - loss: 105.1037 - val_loss: 228.1476\n",
      "Epoch 50/50\n",
      "3523/3523 [==============================] - 44s 12ms/step - loss: 104.0779 - val_loss: 216.0495\n"
     ]
    }
   ],
   "source": [
    "# First Run\n",
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#for name, values in history.history.items():\n",
    "#    plt.plot(values)\n",
    "loss=history.history['loss']\n",
    "val_loss=history.history['val_loss']\n",
    "plt.plot(range(1, len(loss)+1), loss, label='Loss' )\n",
    "plt.plot(range(1, len(loss)+1), val_loss, label='Val Loss' )\n",
    "[plt.axvline(x=i, color='k', lw=.5, ls='--') for i in range(2, 10, 2)]\n",
    "plt.legend()\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_3 (LSTM)                (None, 32)                6144      \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 32)                1056      \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 2)                 66        \n",
      "=================================================================\n",
      "Total params: 9,378\n",
      "Trainable params: 9,378\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Rerun for a smaller epoch\n",
    "model_config = {\"n_epochs\": 8, \"n_neurons\": 64, \"dropout\": 0.4, \"stateful\": False}\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/8\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 104.0128 - val_loss: 217.0459\n",
      "Epoch 2/8\n",
      "3523/3523 [==============================] - 42s 12ms/step - loss: 102.8132 - val_loss: 229.8280\n",
      "Epoch 3/8\n",
      "3523/3523 [==============================] - 45s 13ms/step - loss: 104.2581 - val_loss: 212.4826\n",
      "Epoch 4/8\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 102.4220 - val_loss: 233.0591\n",
      "Epoch 5/8\n",
      "3523/3523 [==============================] - 37s 10ms/step - loss: 102.3404 - val_loss: 212.4230\n",
      "Epoch 6/8\n",
      "3523/3523 [==============================] - 37s 11ms/step - loss: 102.1408 - val_loss: 209.5786\n",
      "Epoch 7/8\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 102.9041 - val_loss: 212.8300\n",
      "Epoch 8/8\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 102.2274 - val_loss: 201.5937\n"
     ]
    }
   ],
   "source": [
    "# First Run\n",
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss=history.history['loss']\n",
    "val_loss=history.history['val_loss']\n",
    "plt.plot(range(1, len(loss)+1), loss, label='Loss' )\n",
    "plt.plot(range(1, len(loss)+1), val_loss, label='Val Loss' )\n",
    "[plt.axvline(x=i, color='k', lw=.5, ls='--') for i in range(2, 8, 2)]\n",
    "plt.legend()\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_ds = timeseries_dataset_from_df(test, data_config[\"batch_size\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "558/558 [==============================] - 3s 5ms/step - loss: 221.3952\n",
      "Test RMSE: 14.88\n"
     ]
    }
   ],
   "source": [
    "test_ds = timeseries_dataset_from_df(test, data_config[\"batch_size\"])\n",
    "mse = model.evaluate(test_ds)\n",
    "print(f\"Test RMSE: {mse**.5:.2f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.8/site-packages/tensorflow/python/training/tracking/tracking.py:111: Model.state_updates (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.8/site-packages/tensorflow/python/training/tracking/tracking.py:111: Layer.updates (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
      "INFO:tensorflow:Assets written to: model/assets\n",
      "{'timesteps': 32, 'batch_size': 32, 'solar_wind_subset': ['bt', 'temperature', 'bx_gse', 'by_gse', 'bz_gse', 'speed', 'density']}\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import pickle\n",
    "\n",
    "model.save(\"model\")\n",
    "\n",
    "with open(\"scaler.pck\", \"wb\") as f:\n",
    "    pickle.dump(scaler, f)\n",
    "\n",
    "data_config[\"solar_wind_subset\"] = SOLAR_WIND_FEATURES\n",
    "print(data_config)\n",
    "with open(\"config.json\", \"w\") as f:\n",
    "    json.dump(data_config, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_5 (LSTM)                (None, 16)                2048      \n",
      "_________________________________________________________________\n",
      "dense_15 (Dense)             (None, 16)                272       \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_16 (Dense)             (None, 2)                 34        \n",
      "=================================================================\n",
      "Total params: 2,354\n",
      "Trainable params: 2,354\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "\n",
    "# define our model\n",
    "model_config = {\"n_epochs\": 50, \"n_neurons\": 16, \"dropout\": 0.25, \"stateful\": False}\n",
    "\n",
    "def make_model3():\n",
    "    model = Sequential()\n",
    "    model.add(\n",
    "        LSTM(\n",
    "            model_config[\"n_neurons\"],\n",
    "            # usually set to (`batch_size`, `sequence_length`, `n_features`)\n",
    "            # setting the batch size to None allows for variable length batches\n",
    "            batch_input_shape=(None, data_config[\"timesteps\"], len(XCOLS)),\n",
    "            stateful=model_config[\"stateful\"],\n",
    "            dropout=model_config[\"dropout\"],\n",
    "        )\n",
    "    )\n",
    "    model.add(Dense(16))\n",
    "    model.add(Dropout(model_config[\"dropout\"]))\n",
    "    \n",
    "    model.add(Dense(len(YCOLS)))\n",
    "    model.compile(\n",
    "        loss=\"mean_squared_error\",\n",
    "        optimizer=\"adam\",\n",
    "    )\n",
    "    return model\n",
    "\n",
    "model = make_model3()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "3523/3523 [==============================] - 48s 14ms/step - loss: 252.5586 - val_loss: 335.9513\n",
      "Epoch 2/50\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 186.6711 - val_loss: 283.5799\n",
      "Epoch 3/50\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 171.4667 - val_loss: 257.2275\n",
      "Epoch 4/50\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 162.2906 - val_loss: 244.6069\n",
      "Epoch 5/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 154.2671 - val_loss: 228.5789\n",
      "Epoch 6/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 150.1504 - val_loss: 216.4760\n",
      "Epoch 7/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 146.1563 - val_loss: 210.6293\n",
      "Epoch 8/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 144.9739 - val_loss: 210.4505\n",
      "Epoch 9/50\n",
      "3523/3523 [==============================] - 38s 11ms/step - loss: 142.4210 - val_loss: 202.8951\n",
      "Epoch 10/50\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 139.4443 - val_loss: 202.0575\n",
      "Epoch 11/50\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 138.0966 - val_loss: 195.6235\n",
      "Epoch 12/50\n",
      "3523/3523 [==============================] - 42s 12ms/step - loss: 137.3370 - val_loss: 192.2656\n",
      "Epoch 13/50\n",
      "3523/3523 [==============================] - 51s 14ms/step - loss: 135.4738 - val_loss: 184.7434\n",
      "Epoch 14/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 134.4939 - val_loss: 192.1328\n",
      "Epoch 15/50\n",
      "3523/3523 [==============================] - 46s 13ms/step - loss: 132.6789 - val_loss: 185.5335\n",
      "Epoch 16/50\n",
      "3523/3523 [==============================] - 38s 11ms/step - loss: 131.9563 - val_loss: 190.6022\n",
      "Epoch 17/50\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 131.2827 - val_loss: 199.0649\n",
      "Epoch 18/50\n",
      "3523/3523 [==============================] - 38s 11ms/step - loss: 129.8580 - val_loss: 206.3409\n",
      "Epoch 19/50\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 129.9276 - val_loss: 195.5734\n",
      "Epoch 20/50\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 128.4144 - val_loss: 213.5421\n",
      "Epoch 21/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 128.9483 - val_loss: 205.3890\n",
      "Epoch 22/50\n",
      "3523/3523 [==============================] - 40s 11ms/step - loss: 127.6262 - val_loss: 201.1602\n",
      "Epoch 23/50\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 127.6733 - val_loss: 208.0570\n",
      "Epoch 24/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 127.2200 - val_loss: 210.3936\n",
      "Epoch 25/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 125.7100 - val_loss: 206.4345\n",
      "Epoch 26/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 126.4926 - val_loss: 224.9608\n",
      "Epoch 27/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 125.8086 - val_loss: 224.8888\n",
      "Epoch 28/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 124.3963 - val_loss: 213.0099\n",
      "Epoch 29/50\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 124.6447 - val_loss: 220.9126\n",
      "Epoch 30/50\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 123.3384 - val_loss: 214.4035\n",
      "Epoch 31/50\n",
      "3523/3523 [==============================] - 39s 11ms/step - loss: 123.8783 - val_loss: 228.2259\n",
      "Epoch 32/50\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 123.3900 - val_loss: 217.3127\n",
      "Epoch 33/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 122.7093 - val_loss: 216.1199\n",
      "Epoch 34/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 123.9574 - val_loss: 212.1627\n",
      "Epoch 35/50\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 121.2319 - val_loss: 211.5317\n",
      "Epoch 36/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 120.9317 - val_loss: 214.2530\n",
      "Epoch 37/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 121.7797 - val_loss: 217.7367\n",
      "Epoch 38/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 120.9890 - val_loss: 213.5626\n",
      "Epoch 39/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 120.1108 - val_loss: 215.4474\n",
      "Epoch 40/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 119.7613 - val_loss: 206.5069\n",
      "Epoch 41/50\n",
      "3523/3523 [==============================] - 33s 9ms/step - loss: 120.7468 - val_loss: 193.8525\n",
      "Epoch 42/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 119.3657 - val_loss: 208.1695\n",
      "Epoch 43/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 120.5190 - val_loss: 206.2519\n",
      "Epoch 44/50\n",
      "3523/3523 [==============================] - 33s 9ms/step - loss: 120.0656 - val_loss: 204.0914\n",
      "Epoch 45/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 118.3577 - val_loss: 205.4373\n",
      "Epoch 46/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 118.1170 - val_loss: 203.9927\n",
      "Epoch 47/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 118.8504 - val_loss: 212.2828\n",
      "Epoch 48/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 117.9547 - val_loss: 211.9092\n",
      "Epoch 49/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 117.8473 - val_loss: 213.2433\n",
      "Epoch 50/50\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 117.4925 - val_loss: 219.1937\n"
     ]
    }
   ],
   "source": [
    "# First Run\n",
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss=history.history['loss']\n",
    "val_loss=history.history['val_loss']\n",
    "plt.plot(range(1, len(loss)+1), loss, label='Loss' )\n",
    "plt.plot(range(1, len(loss)+1), val_loss, label='Val Loss' )\n",
    "[plt.axvline(x=i, color='k', lw=.5, ls='--') for i in range(2, 18, 2)]\n",
    "plt.legend()\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_5 (LSTM)                (None, 16)                2048      \n",
      "_________________________________________________________________\n",
      "dense_15 (Dense)             (None, 16)                272       \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_16 (Dense)             (None, 2)                 34        \n",
      "=================================================================\n",
      "Total params: 2,354\n",
      "Trainable params: 2,354\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Rerun for a smaller epoch\n",
    "model_config = {\"n_epochs\": 17, \"n_neurons\": 16, \"dropout\": 0.25, \"stateful\": False}\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/17\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 117.7067 - val_loss: 219.7196\n",
      "Epoch 2/17\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 118.3973 - val_loss: 222.6059\n",
      "Epoch 3/17\n",
      "3523/3523 [==============================] - 36s 10ms/step - loss: 117.1724 - val_loss: 211.9718\n",
      "Epoch 4/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 116.0797 - val_loss: 214.3005\n",
      "Epoch 5/17\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 116.6488 - val_loss: 211.3860\n",
      "Epoch 6/17\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 115.2719 - val_loss: 207.8973\n",
      "Epoch 7/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 117.6797 - val_loss: 211.0687\n",
      "Epoch 8/17\n",
      "3523/3523 [==============================] - 33s 9ms/step - loss: 116.1691 - val_loss: 231.6855\n",
      "Epoch 9/17\n",
      "3523/3523 [==============================] - 35s 10ms/step - loss: 115.9288 - val_loss: 219.5872\n",
      "Epoch 10/17\n",
      "3523/3523 [==============================] - 33s 9ms/step - loss: 115.9145 - val_loss: 235.8772\n",
      "Epoch 11/17\n",
      "3523/3523 [==============================] - 33s 9ms/step - loss: 115.2618 - val_loss: 229.8589\n",
      "Epoch 12/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 115.5535 - val_loss: 208.2236\n",
      "Epoch 13/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 114.3523 - val_loss: 210.2076\n",
      "Epoch 14/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 115.9448 - val_loss: 206.0572\n",
      "Epoch 15/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 114.4751 - val_loss: 217.0953\n",
      "Epoch 16/17\n",
      "3523/3523 [==============================] - 37s 10ms/step - loss: 115.3824 - val_loss: 221.2699\n",
      "Epoch 17/17\n",
      "3523/3523 [==============================] - 34s 10ms/step - loss: 114.5439 - val_loss: 221.7371\n"
     ]
    }
   ],
   "source": [
    "# First Run\n",
    "history = model.fit(\n",
    "    train_ds,\n",
    "    batch_size=data_config[\"batch_size\"],\n",
    "    epochs=model_config[\"n_epochs\"],\n",
    "    verbose=1,\n",
    "    shuffle=False,\n",
    "    validation_data=val_ds,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss=history.history['loss']\n",
    "val_loss=history.history['val_loss']\n",
    "plt.plot(range(1, len(loss)+1), loss, label='Loss' )\n",
    "plt.plot(range(1, len(loss)+1), val_loss, label='Val Loss' )\n",
    "[plt.axvline(x=i, color='k', lw=.5, ls='--') for i in range(2, 18, 2)]\n",
    "plt.legend()\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "558/558 [==============================] - 2s 4ms/step - loss: 190.0041\n",
      "Test RMSE: 13.78\n"
     ]
    }
   ],
   "source": [
    "test_ds = timeseries_dataset_from_df(test, data_config[\"batch_size\"])\n",
    "mse = model.evaluate(test_ds)\n",
    "print(f\"Test RMSE: {mse**.5:.2f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: model/assets\n",
      "{'timesteps': 32, 'batch_size': 32, 'solar_wind_subset': ['bt', 'temperature', 'bx_gse', 'by_gse', 'bz_gse', 'speed', 'density']}\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import pickle\n",
    "\n",
    "model.save(\"model\")\n",
    "\n",
    "with open(\"scaler.pck\", \"wb\") as f:\n",
    "    pickle.dump(scaler, f)\n",
    "\n",
    "data_config[\"solar_wind_subset\"] = SOLAR_WIND_FEATURES\n",
    "print(data_config)\n",
    "with open(\"config.json\", \"w\") as f:\n",
    "    json.dump(data_config, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import pickle\n",
    "from typing import Tuple\n",
    "from tensorflow import keras\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "# Load in serialized model, config, and scaler\n",
    "model = keras.models.load_model(\"model\")\n",
    "\n",
    "with open(\"config.json\", \"r\") as f:\n",
    "    CONFIG = json.load(f)\n",
    "\n",
    "with open(\"scaler.pck\", \"rb\") as f:\n",
    "    scaler = pickle.load(f)\n",
    "\n",
    "# Set global variables    \n",
    "TIMESTEPS = CONFIG[\"timesteps\"]\n",
    "SOLAR_WIND_FEATURES = [\n",
    "    \"bt\",\n",
    "    \"temperature\",\n",
    "    \"bx_gse\",\n",
    "    \"by_gse\",\n",
    "    \"bz_gse\",\n",
    "    \"speed\",\n",
    "    \"density\",\n",
    "]\n",
    "XCOLS = (\n",
    "    [col + \"_mean\" for col in SOLAR_WIND_FEATURES]\n",
    "    + [col + \"_std\" for col in SOLAR_WIND_FEATURES]\n",
    "    + [\"smoothed_ssn\"]\n",
    ")\n",
    "\n",
    "\n",
    "# Define functions for preprocessing\n",
    "def impute_features(feature_df):\n",
    "    \"\"\"Imputes data using the following methods:\n",
    "    - `smoothed_ssn`: forward fill\n",
    "    - `solar_wind`: interpolation\n",
    "    \"\"\"\n",
    "    # forward fill sunspot data for the rest of the month\n",
    "    feature_df.smoothed_ssn = feature_df.smoothed_ssn.fillna(method=\"ffill\")\n",
    "    # interpolate between missing solar wind values\n",
    "    feature_df = feature_df.interpolate()\n",
    "    return feature_df\n",
    "\n",
    "\n",
    "def aggregate_hourly(feature_df, aggs=[\"mean\", \"std\"]):\n",
    "    \"\"\"Aggregates features to the floor of each hour using mean and standard deviation.\n",
    "    e.g. All values from \"11:00:00\" to \"11:59:00\" will be aggregated to \"11:00:00\".\n",
    "    \"\"\"\n",
    "    # group by the floor of each hour use timedelta index\n",
    "    agged = feature_df.groupby([feature_df.index.floor(\"H\")]).agg(aggs)\n",
    "\n",
    "    # flatten hierachical column index\n",
    "    agged.columns = [\"_\".join(x) for x in agged.columns]\n",
    "    return agged\n",
    "\n",
    "\n",
    "def preprocess_features(solar_wind, sunspots, scaler=None, subset=None):\n",
    "    \"\"\"\n",
    "    Preprocessing steps:\n",
    "        - Subset the data\n",
    "        - Aggregate hourly\n",
    "        - Join solar wind and sunspot data\n",
    "        - Scale using standard scaler\n",
    "        - Impute missing values\n",
    "    \"\"\"\n",
    "    # select features we want to use\n",
    "    if subset:\n",
    "        solar_wind = solar_wind[subset]\n",
    "\n",
    "    # aggregate solar wind data and join with sunspots\n",
    "    hourly_features = aggregate_hourly(solar_wind).join(sunspots)\n",
    "\n",
    "    # subtract mean and divide by standard deviation\n",
    "    if scaler is None:\n",
    "        scaler = StandardScaler()\n",
    "        scaler.fit(hourly_features)\n",
    "\n",
    "    normalized = pd.DataFrame(\n",
    "        scaler.transform(hourly_features),\n",
    "        index=hourly_features.index,\n",
    "        columns=hourly_features.columns,\n",
    "    )\n",
    "\n",
    "    # impute missing values\n",
    "    imputed = impute_features(normalized)\n",
    "\n",
    "    # we want to return the scaler object as well to use later during prediction\n",
    "    return imputed, scaler\n",
    "\n",
    "\n",
    "# THIS MUST BE DEFINED FOR YOUR SUBMISSION TO RUN\n",
    "def predict_dst(\n",
    "    solar_wind_7d: pd.DataFrame,\n",
    "    satellite_positions_7d: pd.DataFrame,\n",
    "    latest_sunspot_number: float,\n",
    ") -> Tuple[float, float]:\n",
    "    \"\"\"\n",
    "    Take all of the data up until time t-1, and then make predictions for\n",
    "    times t and t+1.\n",
    "    Parameters\n",
    "    ----------\n",
    "    solar_wind_7d: pd.DataFrame\n",
    "        The last 7 days of satellite data up until (t - 1) minutes [exclusive of t]\n",
    "    satellite_positions_7d: pd.DataFrame\n",
    "        The last 7 days of satellite position data up until the present time [inclusive of t]\n",
    "    latest_sunspot_number: float\n",
    "        The latest monthly sunspot number (SSN) to be available\n",
    "    Returns\n",
    "    -------\n",
    "    predictions : Tuple[float, float]\n",
    "        A tuple of two predictions, for (t and t + 1 hour) respectively; these should\n",
    "        be between -2,000 and 500.\n",
    "    \"\"\"\n",
    "    # Re-format data to fit into our pipeline\n",
    "    sunspots = pd.DataFrame(index=solar_wind_7d.index, columns=[\"smoothed_ssn\"])\n",
    "    sunspots[\"smoothed_ssn\"] = latest_sunspot_number\n",
    "    \n",
    "    # Process our features and grab last 32 (timesteps) hours\n",
    "    features, s = preprocess_features(\n",
    "        solar_wind_7d, sunspots, scaler=scaler, subset=SOLAR_WIND_FEATURES\n",
    "    )\n",
    "    model_input = features[-TIMESTEPS:][XCOLS].values.reshape(\n",
    "        (1, TIMESTEPS, features.shape[1])\n",
    "    )\n",
    "    \n",
    "    # Make a prediction\n",
    "    prediction_at_t0, prediction_at_t1 = model.predict(model_input)[0]\n",
    "\n",
    "    # Optional check for unexpected values\n",
    "    if not np.isfinite(prediction_at_t0):\n",
    "        prediction_at_t0 = -12\n",
    "    if not np.isfinite(prediction_at_t1):\n",
    "        prediction_at_t1 = -12\n",
    "\n",
    "    return prediction_at_t0, prediction_at_t1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}