8 - Crossvalidation of the Hyperlocal Models.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Crossvalidation of the Hyperlocal Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "sFRp0QrlQuLw"
   },
   "outputs": [],
   "source": [
    "# !pip3 install simdkalman fbprophet scipy==1.2 statsmodels --upgrade"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7ie9EoX2Qx2X"
   },
   "outputs": [],
   "source": [
    "import math\n",
    "import datetime\n",
    "import random\n",
    "import time\n",
    "import multiprocessing\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import simdkalman\n",
    "\n",
    "import statsmodels.api as sm\n",
    "from statsmodels.tsa.holtwinters import SimpleExpSmoothing, ExponentialSmoothing\n",
    "\n",
    "from sklearn.linear_model import LinearRegression, Ridge, HuberRegressor\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "from fbprophet import Prophet\n",
    "from fbprophet.diagnostics import cross_validation, performance_metrics\n",
    "from fbprophet.plot import plot_cross_validation_metric\n",
    "\n",
    "import tensorflow.keras as tfk\n",
    "\n",
    "import _pickle as cPickle\n",
    "\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib import animation\n",
    "sns.set(style=\"whitegrid\")\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "9Mnnx2rURUxu"
   },
   "source": [
    "## The Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "NNiKTC-NRWO5"
   },
   "source": [
    "Load the preprocessed data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !gsutil cp gs://franzfiles/full_data.csv ."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 122
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 29273,
     "status": "ok",
     "timestamp": 1560078258425,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "Ra89js2ZQ0D-",
    "outputId": "6b10e721-7a10-4eb1-aa4f-b742bf1c70af"
   },
   "outputs": [],
   "source": [
    "# from google.colab import drive\n",
    "# drive.mount('/content/gdrive')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 30106,
     "status": "ok",
     "timestamp": 1560078259643,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "xwYgMSETRYZT",
    "outputId": "b69cda4a-91e5-4b02-f59f-a9b255c6a808"
   },
   "outputs": [],
   "source": [
    "# cd '/content/gdrive/My Drive/Colab Notebooks/aiforsea'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 204
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 49785,
     "status": "ok",
     "timestamp": 1560078279594,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "UNjXqe3BReP1",
    "outputId": "8857f87c-fe6e-442c-9aed-172933d20bf1"
   },
   "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>geohash6</th>\n",
       "      <th>day</th>\n",
       "      <th>demand</th>\n",
       "      <th>in_raw</th>\n",
       "      <th>weekday</th>\n",
       "      <th>hour</th>\n",
       "      <th>minutes</th>\n",
       "      <th>ds</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>qp02yc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.653687</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>qp02z1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.664673</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>qp02z3</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.675659</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>qp02z9</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.686646</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>qp02zc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.697632</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  geohash6  day  demand  in_raw  weekday  hour  minutes                   ds  \\\n",
       "0   qp02yc    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "1   qp02z1    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "2   qp02z3    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "3   qp02z9    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "4   qp02zc    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "\n",
       "   latitude  longitude  \n",
       "0 -5.484924  90.653687  \n",
       "1 -5.484924  90.664673  \n",
       "2 -5.484924  90.675659  \n",
       "3 -5.484924  90.686646  \n",
       "4 -5.484924  90.697632  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('full_data.csv')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "sMc-JdHCRhPJ"
   },
   "source": [
    "### Add info for fbprophet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# mask = np.load('temporal_anomaly_detection/density_mask.npy')\n",
    "# flat_mask = mask.reshape(-1).astype('float32')\n",
    "# mask_sum = mask.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def scaled_loss(inputs, outputs):\n",
    "#   outputs *= flat_mask\n",
    "#   loss = tfk.metrics.mean_squared_error(inputs, outputs)\n",
    "#   return mask_sum * loss\n",
    "# \n",
    "# autoencoder = tfk.models.load_model('temporal_anomaly_detection/autoencoder.h5',\n",
    "#                                     custom_objects={'scaled_loss':scaled_loss})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "mahalanobis_dist = np.load('temporal_anomaly_detection/mahalanobis_dist.npy')\n",
    "# train_mu  = np.load('temporal_anomaly_detection/train_mu.npy')\n",
    "# train_std = np.load('temporal_anomaly_detection/train_std.npy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['is_anomaly'] = (df['day'] - 1)*4*24 + df['hour']*4 + df['minutes']//15\n",
    "df['is_anomaly'] = df['is_anomaly'].apply(lambda x : mahalanobis_dist[x] > 4.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1041,
     "status": "ok",
     "timestamp": 1560078292899,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "QHpgUqVMRiZp",
    "outputId": "aeb08a8f-1538-4c25-ecd8-6494504ee5f7"
   },
   "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>geohash6</th>\n",
       "      <th>day</th>\n",
       "      <th>demand</th>\n",
       "      <th>in_raw</th>\n",
       "      <th>weekday</th>\n",
       "      <th>hour</th>\n",
       "      <th>minutes</th>\n",
       "      <th>ds</th>\n",
       "      <th>latitude</th>\n",
       "      <th>longitude</th>\n",
       "      <th>is_anomaly</th>\n",
       "      <th>is_weekend</th>\n",
       "      <th>is_weekday</th>\n",
       "      <th>cap</th>\n",
       "      <th>floor</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>qp02yc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.653687</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>qp02z1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.664673</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>qp02z3</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.675659</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>qp02z9</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.686646</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>qp02zc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>False</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1900-01-01 00:00:00</td>\n",
       "      <td>-5.484924</td>\n",
       "      <td>90.697632</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  geohash6  day  demand  in_raw  weekday  hour  minutes                   ds  \\\n",
       "0   qp02yc    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "1   qp02z1    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "2   qp02z3    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "3   qp02z9    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "4   qp02zc    1     0.0   False        1     0        0  1900-01-01 00:00:00   \n",
       "\n",
       "   latitude  longitude  is_anomaly  is_weekend  is_weekday  cap  floor  \n",
       "0 -5.484924  90.653687       False       False        True  1.0    0.0  \n",
       "1 -5.484924  90.664673       False       False        True  1.0    0.0  \n",
       "2 -5.484924  90.675659       False       False        True  1.0    0.0  \n",
       "3 -5.484924  90.686646       False       False        True  1.0    0.0  \n",
       "4 -5.484924  90.697632       False       False        True  1.0    0.0  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['is_weekend'] = (4 <= (df['day'].values % 7 )) & ((df['day'].values % 7) <= 5)\n",
    "df['is_weekday'] = ~(df['is_weekend'])\n",
    "df['cap']   = 1.0\n",
    "df['floor'] = 0.0\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>geohash6</th>\n",
       "      <th>day</th>\n",
       "      <th>y</th>\n",
       "      <th>ds</th>\n",
       "      <th>is_anomaly</th>\n",
       "      <th>is_weekend</th>\n",
       "      <th>is_weekday</th>\n",
       "      <th>cap</th>\n",
       "      <th>floor</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>qp02yc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1900-01-01</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>qp02z1</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1900-01-01</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>qp02z3</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1900-01-01</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>qp02z9</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1900-01-01</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>qp02zc</td>\n",
       "      <td>1</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1900-01-01</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  geohash6  day    y         ds  is_anomaly  is_weekend  is_weekday  cap  \\\n",
       "0   qp02yc    1  0.0 1900-01-01       False       False        True  1.0   \n",
       "1   qp02z1    1  0.0 1900-01-01       False       False        True  1.0   \n",
       "2   qp02z3    1  0.0 1900-01-01       False       False        True  1.0   \n",
       "3   qp02z9    1  0.0 1900-01-01       False       False        True  1.0   \n",
       "4   qp02zc    1  0.0 1900-01-01       False       False        True  1.0   \n",
       "\n",
       "   floor  \n",
       "0    0.0  \n",
       "1    0.0  \n",
       "2    0.0  \n",
       "3    0.0  \n",
       "4    0.0  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = df.drop(['in_raw', 'weekday', 'hour', 'minutes', 'latitude', 'longitude'], axis=1)\n",
    "df = df.rename(columns={'demand' : 'y'})\n",
    "df['ds'] = pd.to_datetime(df['ds'])\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "3sAPce7nzl3O"
   },
   "source": [
    "### Collect info for each location"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 1486,
     "status": "ok",
     "timestamp": 1560078296010,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "pTGenbdLzpOG",
    "outputId": "84172e20-7296-4134-f139-4047ac627c16"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1329 ['qp02yc' 'qp02z1' 'qp02z3' ... 'qp0djw' 'qp0djy' 'qp0dnn']\n"
     ]
    }
   ],
   "source": [
    "locations = df['geohash6'].unique()\n",
    "print(len(locations), locations)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 18.2 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "gb = df.groupby('geohash6')    \n",
    "location_series = {}\n",
    "for x in gb.groups:\n",
    "    group = gb.get_group(x).reset_index().drop('index', axis=1)\n",
    "    location_series[x] = group"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "pqpfle8HSh7n"
   },
   "source": [
    "## Model Selection"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Define training horizon and phase"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vLxzSVAaaaZP"
   },
   "outputs": [],
   "source": [
    "train_start = pd.to_datetime('1900-01-20 00:00:00')\n",
    "train_stepdelta = datetime.timedelta(days=1, hours=1, minutes=15)\n",
    "train_horizon = datetime.timedelta(days=14)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "TIMESTEPS = 20"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Select locations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "locations_subset = np.random.choice(locations, size=len(locations)//10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test on a location"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fbprophet.forecaster.Prophet at 0x18b85620cc0>"
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "series = location_series['qp09d9']\n",
    "# forecast using fbprophet's default settings\n",
    "fbm1 = Prophet(seasonality_mode='multiplicative',\n",
    "               weekly_seasonality=True,\n",
    "               yearly_seasonality=False)\n",
    "fbm1.add_regressor('is_anomaly')\n",
    "fbm1.fit(series)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "future = fbm1.make_future_dataframe(periods=5, freq='15min', include_history=True)\n",
    "future['is_anomaly'] = np.concatenate((series['is_anomaly'].values, np.zeros(5)), axis=None)\n",
    "forecast1 = fbm1.predict(future)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 648x864 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig2 = fbm1.plot_components(forecast1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Crossvalidate the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def worker(timestep, df_data, return_dict):\n",
    "    # train-test split\n",
    "    df_train = df_data.iloc[:-5]\n",
    "    df_test  = df_data.iloc[-5:]\n",
    "    X_train = np.arange(len(df_train)).reshape(-1, 1)\n",
    "    X_test  = (np.arange(5) + len(df_train)).reshape(-1, 1)\n",
    "\n",
    "    # forecast using data from previous cycle\n",
    "    prev_pred = df_train['y'].values[-4*24: -4*24+5]\n",
    "\n",
    "    # forecast using Keano Reeves\n",
    "    kf = simdkalman.KalmanFilter(\n",
    "      state_transition = np.array([[1,1],[0,1]]),\n",
    "      process_noise = np.diag([0.1, 0.01]),\n",
    "      observation_model = np.array([[1,0]]),\n",
    "      observation_noise = 0.5)\n",
    "    kalman_pred = kf.predict(df_train['y'].values, 5).states.mean[:,0]\n",
    "\n",
    "    # deseasonalize\n",
    "    decomposition = sm.tsa.seasonal_decompose(df_train['y'] + 1e-9,\n",
    "                                            model='multiplicative',\n",
    "                                            freq=4*24,\n",
    "                                            two_sided=False,\n",
    "                                            extrapolate_trend=1)\n",
    "    df_train['deseasonalized'] = df_train['y'].values / (decomposition.seasonal + 1e-9)\n",
    "    seasonal_test = decomposition.seasonal.values[-4*24:-4*24+5]\n",
    "\n",
    "    # forecast using theta = 0 (with linear regression)\n",
    "    linreg = HuberRegressor().fit(X_train, df_train['deseasonalized'].values)\n",
    "    theta0_pred = linreg.predict(X_test)\n",
    "    y0_pred = theta0_pred * seasonal_test\n",
    "\n",
    "    # forecast using theta = 2 (with simple exponential smoothing)\n",
    "    y_theta2 = 2*df_train['deseasonalized'].values - linreg.predict(X_train)\n",
    "    ses = SimpleExpSmoothing(y_theta2).fit()\n",
    "    theta2_pred = ses.forecast(5)\n",
    "    y2_pred = theta2_pred * seasonal_test\n",
    "\n",
    "    # forecast using fbprophet's default settings\n",
    "    fbm1 = Prophet(seasonality_mode='multiplicative',\n",
    "                   weekly_seasonality=True,\n",
    "                   yearly_seasonality=False)\n",
    "    fbm1.add_regressor('is_anomaly')\n",
    "    fbm1.fit(df_train)\n",
    "    \n",
    "    future = fbm1.make_future_dataframe(periods=5, freq='15min', include_history=False)\n",
    "    future['is_anomaly'] = 0\n",
    "    forecast1 = fbm1.predict(future)\n",
    "\n",
    "\n",
    "    # combinations\n",
    "    # all_pred            = (prev_pred + kalman_pred + y0_pred + y2_pred + forecast1['yhat'].values) / 5.0\n",
    "\n",
    "    # clip\n",
    "    # prev_pred           = np.clip(prev_pred               , a_min=0.0, a_max=1.0)\n",
    "    kalman_pred         = np.clip(kalman_pred             , a_min=0.0, a_max=1.0)\n",
    "    y0_pred             = np.clip(y0_pred                 , a_min=0.0, a_max=1.0)\n",
    "    y2_pred             = np.clip(y2_pred                 , a_min=0.0, a_max=1.0)\n",
    "    forecast1['yhat']   = np.clip(forecast1['yhat'].values, a_min=0.0, a_max=1.0)\n",
    "    # all_pred            = np.clip(all_pred                , a_min=0.0, a_max=1.0)\n",
    "\n",
    "    # save\n",
    "    predictions = np.asarray([prev_pred,\n",
    "                              kalman_pred,\n",
    "                              y0_pred,\n",
    "                              y2_pred,\n",
    "                              forecast1['yhat'].values])\n",
    "    actuals = df_test['y'].values\n",
    "    \n",
    "    return_dict[timestep] = {'predictions' : predictions,\n",
    "                             'actuals'     : actuals}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1900-01-20 00:00:00 1900-02-03 00:00:00\n",
      "1900-01-21 01:15:00 1900-02-04 01:15:00\n",
      "1900-01-22 02:30:00 1900-02-05 02:30:00\n",
      "1900-01-23 03:45:00 1900-02-06 03:45:00\n",
      "1900-01-24 05:00:00 1900-02-07 05:00:00\n",
      "1900-01-25 06:15:00 1900-02-08 06:15:00\n",
      "1900-01-26 07:30:00 1900-02-09 07:30:00\n",
      "1900-01-27 08:45:00 1900-02-10 08:45:00\n",
      "1900-01-28 10:00:00 1900-02-11 10:00:00\n",
      "1900-01-29 11:15:00 1900-02-12 11:15:00\n",
      "1900-01-30 12:30:00 1900-02-13 12:30:00\n",
      "1900-01-31 13:45:00 1900-02-14 13:45:00\n",
      "1900-02-01 15:00:00 1900-02-15 15:00:00\n",
      "1900-02-02 16:15:00 1900-02-16 16:15:00\n",
      "1900-02-03 17:30:00 1900-02-17 17:30:00\n",
      "1900-02-04 18:45:00 1900-02-18 18:45:00\n",
      "1900-02-05 20:00:00 1900-02-19 20:00:00\n",
      "1900-02-06 21:15:00 1900-02-20 21:15:00\n",
      "1900-02-07 22:30:00 1900-02-21 22:30:00\n",
      "1900-02-08 23:45:00 1900-02-22 23:45:00\n",
      "(1, 20, 5, 5) (1, 20, 5)\n",
      "Wall time: 2min 11s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "predictions = []\n",
    "actuals     = []\n",
    "for location in ['qp02zw']:\n",
    "    # location = locations_subset[iloc]\n",
    "    # print(\"{}/{} - currently processing location {}\".format(iloc+1, len(locations_subset), location))\n",
    "    \n",
    "    manager = multiprocessing.Manager()\n",
    "    return_dict = manager.dict()\n",
    "    jobs = []\n",
    "    for i in range(TIMESTEPS):      \n",
    "        start_date = train_start + i*train_stepdelta\n",
    "        end_date   = start_date + train_horizon\n",
    "        print(start_date, end_date)\n",
    "        \n",
    "        df_loc = location_series[location]\n",
    "        df_loc = df_loc[(start_date <= df_loc['ds']) & (df_loc['ds'] <= end_date)]\n",
    "        df_loc = df_loc.reset_index().drop('index', axis=1)\n",
    "        worker(i, df_loc, return_dict)\n",
    "#         \n",
    "#         df_loc = location_series[location]\n",
    "#         df_loc = df_loc[(start_date <= df_loc['ds']) & (df_loc['ds'] <= end_date)]\n",
    "#         df_loc = df_loc.reset_index().drop('index', axis=1)\n",
    "#         \n",
    "#         p = multiprocessing.Process(target=worker, args=(i, df_loc, return_dict))\n",
    "#         jobs.append(p)\n",
    "#         p.start()\n",
    "# \n",
    "#     for proc in jobs:\n",
    "#         proc.join()\n",
    "\n",
    "    predictions.append(np.stack([return_dict[i]['predictions'] for i in range(TIMESTEPS)], axis=0))\n",
    "    actuals.append(np.stack([return_dict[i]['actuals']     for i in range(TIMESTEPS)], axis=0))\n",
    "\n",
    "predictions = np.asarray(predictions)\n",
    "actuals     = np.asarray(actuals)\n",
    "print(predictions.shape, actuals.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "prev_mse_sum           = 0\n",
    "kalman_mse_sum         = 0\n",
    "\n",
    "y0_mse_sum             = 0\n",
    "y2_mse_sum             = 0\n",
    "theta_mse_sum          = 0\n",
    "\n",
    "fbm1_mse_sum           = 0\n",
    "\n",
    "all_mse_sum            = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(len(predictions)):\n",
    "    for t in range(TIMESTEPS):\n",
    "        prev_mse_sum   += mean_squared_error(predictions[i][t][0], actuals[i][t])\n",
    "        kalman_mse_sum += mean_squared_error(predictions[i][t][1], actuals[i][t])\n",
    "        y0_mse_sum     += mean_squared_error(predictions[i][t][2], actuals[i][t])\n",
    "        y2_mse_sum     += mean_squared_error(predictions[i][t][3], actuals[i][t])\n",
    "        fbm1_mse_sum   += mean_squared_error(predictions[i][t][4], actuals[i][t])\n",
    "        \n",
    "        theta_pred = (predictions[i][t][2] + predictions[i][t][3]) / 2.0\n",
    "        theta_pred = np.clip(theta_pred, a_min=0.0, a_max=1.0)\n",
    "        theta_mse_sum  += mean_squared_error(theta_pred, actuals[i][t])\n",
    "        \n",
    "        # all_pred   = predictions[i][t].mean(axis=0)\n",
    "        all_pred = (predictions[i][t][0] + predictions[i][t][1] + theta_pred + predictions[i][t][4]) / 4.0\n",
    "        all_mse_sum    += mean_squared_error(all_pred, actuals[i][t])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 204
    },
    "colab_type": "code",
    "executionInfo": {
     "elapsed": 176636,
     "status": "ok",
     "timestamp": 1560079636633,
     "user": {
      "displayName": "Franz Louis Cesista",
      "photoUrl": "https://lh3.googleusercontent.com/-NquKeLt1770/AAAAAAAAAAI/AAAAAAAAAMo/RHnx0T1cGIk/s64/photo.jpg",
      "userId": "08273419339046628946"
     },
     "user_tz": -480
    },
    "id": "aWwIYT5tbLqE",
    "outputId": "18fcc4ae-b0b6-4eff-c031-f05fc5292aad"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "prev_rmse            0.07980296192739604\n",
      "kalman_rmse          0.09829175225854198\n",
      "y0_rmse              0.08463593714537263\n",
      "y2_rmse              0.1157874603786236\n",
      "fbm1_rmse            0.06820627883046501\n",
      "theta_rmse           0.07462596287235407\n",
      "all_rmse             0.05649223125013117\n"
     ]
    }
   ],
   "source": [
    "print(\"prev_rmse           \", math.sqrt(prev_mse_sum          /(TIMESTEPS * len(predictions))))\n",
    "print(\"kalman_rmse         \", math.sqrt(kalman_mse_sum        /(TIMESTEPS * len(predictions))))\n",
    "print(\"y0_rmse             \", math.sqrt(y0_mse_sum            /(TIMESTEPS * len(predictions))))\n",
    "print(\"y2_rmse             \", math.sqrt(y2_mse_sum            /(TIMESTEPS * len(predictions))))\n",
    "print(\"fbm1_rmse           \", math.sqrt(fbm1_mse_sum          /(TIMESTEPS * len(predictions))))\n",
    "# print(\"fbm2_rmse           \", math.sqrt(fbm2_mse_sum          /(TIMESTEPS * len(predictions))))\n",
    "# print(\"theta_fb1_rmse      \", math.sqrt(theta_fb1_mse_sum     /(TIMESTEPS * len(predictions))))\n",
    "# print(\"theta_fb2_rmse      \", math.sqrt(theta_fb2_mse_sum     /(TIMESTEPS * len(predictions))))\n",
    "# print(\"theta_fb1_prev_rmse \", math.sqrt(theta_fb1_prev_mse_sum/(TIMESTEPS * len(predictions))))\n",
    "# print(\"theta_fb2_prev_rmse \", math.sqrt(theta_fb2_prev_mse_sum/(TIMESTEPS * len(predictions))))\n",
    "print(\"theta_rmse          \", math.sqrt(theta_mse_sum         /(TIMESTEPS * len(predictions))))\n",
    "print(\"all_rmse            \", math.sqrt(all_mse_sum           /(TIMESTEPS * len(predictions))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "ename": "SyntaxError",
     "evalue": "invalid syntax (<ipython-input-25-6b57b8d4bbe1>, line 1)",
     "output_type": "error",
     "traceback": [
      "\u001b[1;36m  File \u001b[1;32m\"<ipython-input-25-6b57b8d4bbe1>\"\u001b[1;36m, line \u001b[1;32m1\u001b[0m\n\u001b[1;33m    prev_rmse            0.07980296192739604\u001b[0m\n\u001b[1;37m                                           ^\u001b[0m\n\u001b[1;31mSyntaxError\u001b[0m\u001b[1;31m:\u001b[0m invalid syntax\n"
     ]
    }
   ],
   "source": [
    "prev_rmse            0.07980296192739604\n",
    "kalman_rmse          0.09829175225854198\n",
    "y0_rmse              0.08463594529766318\n",
    "y2_rmse              0.11578745803879455\n",
    "fbm1_rmse            0.0685048706433476\n",
    "theta_rmse           0.07462596399847413\n",
    "all_rmse             0.05658998701722148"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Save "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "crW0fL2acFB8"
   },
   "outputs": [],
   "source": [
    "np.save('predictions.npy', predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.save('actuals.npy', actuals)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.save('locations_subset.npy', locations_subset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "name": "6.5 - Theta + FBProphet Model Selection (unfinished).ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}