rnn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import pandas as pd\n",
    "from pathlib import Path\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/jovyan/work\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "print(os.getcwd())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "folder_path = 'data/4'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def from_json_folder_to_json_docs(folder_path):\n",
    "    pathlist = Path(folder_path).glob('**/*.json')\n",
    "    json_docs = []\n",
    "    for path in pathlist:\n",
    "        # because path is object not string\n",
    "        path_in_str = str(path)\n",
    "        print('processing file: ' + path_in_str)\n",
    "        with open(path_in_str, encoding=\"utf8\") as f:\n",
    "            json_file = json.load(f)\n",
    "            json_docs.append(json_file)\n",
    "    print('Done!')\n",
    "    return json_docs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "processing file: data/4/1-10.json\n",
      "processing file: data/4/11-20.json\n",
      "processing file: data/4/21-30.json\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "json_data = from_json_folder_to_json_docs(folder_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "first_it = True\n",
    "for json in json_data:\n",
    "    #print(json.keys())\n",
    "    if first_it != True:\n",
    "        alerts_df.append(pd.DataFrame(json['alerts']), ignore_index=True)\n",
    "        jams_df.append(pd.DataFrame(json['jams']), ignore_index=True)\n",
    "    else:\n",
    "        alerts_df = pd.DataFrame(json['alerts'])\n",
    "        jams_df = pd.DataFrame(json['jams'])\n",
    "        first_it = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "del json_data"
   ]
  },
  {
   "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>uuid</th>\n",
       "      <th>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>country</th>\n",
       "      <th>report_rating</th>\n",
       "      <th>confidence</th>\n",
       "      <th>reliability</th>\n",
       "      <th>magvar</th>\n",
       "      <th>location</th>\n",
       "      <th>publish_date</th>\n",
       "      <th>last_seen</th>\n",
       "      <th>city</th>\n",
       "      <th>road_type</th>\n",
       "      <th>street</th>\n",
       "      <th>report_description</th>\n",
       "      <th>thumbs_up</th>\n",
       "      <th>report_by_partner</th>\n",
       "      <th>jam_uuid</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>8271e6bf-fca0-3cad-bac4-6a6959d16b1c</td>\n",
       "      <td>4</td>\n",
       "      <td>13.0</td>\n",
       "      <td>CI</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>10</td>\n",
       "      <td>265</td>\n",
       "      <td>[-70.646227, -33.635035]</td>\n",
       "      <td>2018-04-01T01:02:02+00:00</td>\n",
       "      <td>2018-04-01T03:00:21+00:00</td>\n",
       "      <td>None</td>\n",
       "      <td>3.0</td>\n",
       "      <td>Ruta 79</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>e7b8841f-a144-3b7d-b685-05bc447b73fd</td>\n",
       "      <td>3</td>\n",
       "      <td>4.0</td>\n",
       "      <td>CI</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>56</td>\n",
       "      <td>[-70.670608, -33.398605]</td>\n",
       "      <td>2018-04-01T02:30:58+00:00</td>\n",
       "      <td>2018-04-01T03:00:21+00:00</td>\n",
       "      <td>Conchalí</td>\n",
       "      <td>2.0</td>\n",
       "      <td>Av. Dorsal</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>e1c149d5-4383-3490-bb2d-550156e3ad51</td>\n",
       "      <td>3</td>\n",
       "      <td>3.0</td>\n",
       "      <td>CI</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>[-70.575705, -33.469535]</td>\n",
       "      <td>2018-04-01T02:31:00+00:00</td>\n",
       "      <td>2018-04-01T03:00:21+00:00</td>\n",
       "      <td>Ñuñoa</td>\n",
       "      <td>2.0</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>0</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                   uuid  alert_type_id  alert_subtype_id  \\\n",
       "0  8271e6bf-fca0-3cad-bac4-6a6959d16b1c              4              13.0   \n",
       "1  e7b8841f-a144-3b7d-b685-05bc447b73fd              3               4.0   \n",
       "2  e1c149d5-4383-3490-bb2d-550156e3ad51              3               3.0   \n",
       "\n",
       "  country  report_rating  confidence  reliability  magvar  \\\n",
       "0      CI              0           5           10     265   \n",
       "1      CI              1           0            5      56   \n",
       "2      CI              1           0            5       2   \n",
       "\n",
       "                   location               publish_date  \\\n",
       "0  [-70.646227, -33.635035]  2018-04-01T01:02:02+00:00   \n",
       "1  [-70.670608, -33.398605]  2018-04-01T02:30:58+00:00   \n",
       "2  [-70.575705, -33.469535]  2018-04-01T02:31:00+00:00   \n",
       "\n",
       "                   last_seen      city  road_type      street  \\\n",
       "0  2018-04-01T03:00:21+00:00      None        3.0     Ruta 79   \n",
       "1  2018-04-01T03:00:21+00:00  Conchalí        2.0  Av. Dorsal   \n",
       "2  2018-04-01T03:00:21+00:00     Ñuñoa        2.0        None   \n",
       "\n",
       "  report_description  thumbs_up report_by_partner jam_uuid  \n",
       "0               None          0              None     None  \n",
       "1               None          0              None     None  \n",
       "2               None          0              None     None  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(115547, 18)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# parse date to datetime object\n",
    "alerts_df['publish_date_dt'] = pd.to_datetime(alerts_df['publish_date'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# subsets dataframe\n",
    "alerts_df.drop(alerts_df.columns.difference(['alert_subtype_id', 'alert_type_id', 'street', 'publish_date_dt']), 1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop duplicates\n",
    "alerts_df.drop_duplicates(inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop nans from alert_subtype_id, street\n",
    "alerts_df.dropna(subset=['alert_subtype_id','street'], inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# sort by date just in case\n",
    "alerts_df.sort_values(by=['publish_date_dt'], inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set date as index, not sure if good idea\n",
    "#alerts_df = alerts_df.set_index('publish_date_dt').sort_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>115496</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Alberto Llona</td>\n",
       "      <td>2016-08-06 17:55:23+00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115542</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Manuel Antonio Matta</td>\n",
       "      <td>2017-01-14 15:01:37+00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115497</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Manuel Antonio Matta</td>\n",
       "      <td>2017-01-14 16:01:37+00:00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        alert_type_id  alert_subtype_id                street  \\\n",
       "115496              4               7.0         Alberto Llona   \n",
       "115542              4               7.0  Manuel Antonio Matta   \n",
       "115497              4               7.0  Manuel Antonio Matta   \n",
       "\n",
       "                 publish_date_dt  \n",
       "115496 2016-08-06 17:55:23+00:00  \n",
       "115542 2017-01-14 15:01:37+00:00  \n",
       "115497 2017-01-14 16:01:37+00:00  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df.head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(106885, 4)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f08f9840940>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "alerts_df['alert_type_id'].value_counts().plot(kind='barh')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f08f9811fd0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "alerts_df['alert_subtype_id'][alerts_df['alert_type_id'] == 4].value_counts().plot(kind='barh')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 1.  2.  3.  4.  5.  7.  8.  9. 10. 11. 12. 13. 14. 15. 16. 17. 20. 28.\n",
      " 29. 30. 31. 32. 33. 34.]\n"
     ]
    }
   ],
   "source": [
    "print(np.sort(alerts_df['alert_subtype_id'].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1. 2.]\n"
     ]
    }
   ],
   "source": [
    "print(np.sort(alerts_df['alert_subtype_id'][alerts_df['alert_type_id'] == 2].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[3. 4. 5.]\n"
     ]
    }
   ],
   "source": [
    "print(np.sort(alerts_df['alert_subtype_id'][alerts_df['alert_type_id'] == 3].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 7.  8.  9. 10. 11. 12. 13. 14. 15. 16. 17. 20. 28. 29. 30. 34.]\n"
     ]
    }
   ],
   "source": [
    "print(np.sort(alerts_df['alert_subtype_id'][alerts_df['alert_type_id'] == 4].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[31. 32. 33.]\n"
     ]
    }
   ],
   "source": [
    "print(np.sort(alerts_df['alert_subtype_id'][alerts_df['alert_type_id'] == 7].unique()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I have to assume that 'alert_type_id' is 3 (JAM) and 'alert_subtype_id' 4 is JAM_HEAVY_TRAFFIC and 5 is JAM_STAND_STILL_TRAFFIC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3140,)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df['street'].unique().shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# res_df = alerts_df[:200].apply(lambda row: list(alerts_df['alert_subtype_id'][(alerts_df['street'] == row['street']) & \n",
    "#                                                                (alerts_df['publish_date_dt'] <= row['publish_date_dt']) & \n",
    "#                                                                (alerts_df['publish_date_dt'] >= row['publish_date_dt'] - pd.Timedelta(minutes=30))])\n",
    "#                , axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "import dask.dataframe as dd\n",
    "from dask.multiprocessing import get\n",
    "from dask.distributed import Client, LocalCluster\n",
    "cluster = LocalCluster(n_workers=4, threads_per_worker=2)\n",
    "client = Client(cluster, asynchronous=True,\n",
    "                memory_limit='8GB')\n",
    "\n",
    "ddata = dd.from_pandas(alerts_df, npartitions=36)\n",
    "\n",
    "alerts_df['seq'] = ddata.map_partitions(lambda df: df.apply((lambda row: list(alerts_df['alert_subtype_id'][(alerts_df['street'] == row['street']) & \n",
    "                                                               (alerts_df['publish_date_dt'] < row['publish_date_dt']) & \n",
    "                                                               (alerts_df['publish_date_dt'] >= row['publish_date_dt'] - pd.Timedelta('00:30:00'))]))\n",
    "                                                , axis=1)).compute(scheduler=client)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "del ddata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "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>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "      <th>seq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>115496</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Alberto Llona</td>\n",
       "      <td>2016-08-06 17:55:23+00:00</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115542</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Manuel Antonio Matta</td>\n",
       "      <td>2017-01-14 15:01:37+00:00</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115497</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Manuel Antonio Matta</td>\n",
       "      <td>2017-01-14 16:01:37+00:00</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115543</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Av. Independencia</td>\n",
       "      <td>2017-02-14 12:40:23+00:00</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115498</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Av. Independencia</td>\n",
       "      <td>2017-02-14 13:40:23+00:00</td>\n",
       "      <td>[]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        alert_type_id  alert_subtype_id                street  \\\n",
       "115496              4               7.0         Alberto Llona   \n",
       "115542              4               7.0  Manuel Antonio Matta   \n",
       "115497              4               7.0  Manuel Antonio Matta   \n",
       "115543              4               7.0     Av. Independencia   \n",
       "115498              4               7.0     Av. Independencia   \n",
       "\n",
       "                 publish_date_dt seq  \n",
       "115496 2016-08-06 17:55:23+00:00  []  \n",
       "115542 2017-01-14 15:01:37+00:00  []  \n",
       "115497 2017-01-14 16:01:37+00:00  []  \n",
       "115543 2017-02-14 12:40:23+00:00  []  \n",
       "115498 2017-02-14 13:40:23+00:00  []  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "alerts_df_w_seq = alerts_df[alerts_df.astype(str)['seq'] != '[]'].copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "      <th>seq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>115494</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:16+00:00</td>\n",
       "      <td>[33.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115508</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:17+00:00</td>\n",
       "      <td>[33.0, 33.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115495</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:18+00:00</td>\n",
       "      <td>[33.0, 33.0, 33.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115478</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Carlos Silva Vildósola</td>\n",
       "      <td>2018-03-05 15:37:45+00:00</td>\n",
       "      <td>[33.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>4</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Autopista Central</td>\n",
       "      <td>2018-04-01 02:54:35+00:00</td>\n",
       "      <td>[3.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        alert_type_id  alert_subtype_id                  street  \\\n",
       "115494              7              33.0           Av. O'Higgins   \n",
       "115508              7              33.0           Av. O'Higgins   \n",
       "115495              7              33.0           Av. O'Higgins   \n",
       "115478              7              33.0  Carlos Silva Vildósola   \n",
       "27                  4               7.0       Autopista Central   \n",
       "\n",
       "                 publish_date_dt                 seq  \n",
       "115494 2018-02-22 18:55:16+00:00              [33.0]  \n",
       "115508 2018-02-22 18:55:17+00:00        [33.0, 33.0]  \n",
       "115495 2018-02-22 18:55:18+00:00  [33.0, 33.0, 33.0]  \n",
       "115478 2018-03-05 15:37:45+00:00              [33.0]  \n",
       "27     2018-04-01 02:54:35+00:00               [3.0]  "
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(72528, 5)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "      <th>seq</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>3</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Antonia López de Bello</td>\n",
       "      <td>2018-04-01 04:07:34+00:00</td>\n",
       "      <td>[4.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78</th>\n",
       "      <td>3</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Antonia López de Bello</td>\n",
       "      <td>2018-04-01 04:56:23+00:00</td>\n",
       "      <td>[4.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>177</th>\n",
       "      <td>3</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Boulevard Lyon Juana de Arco</td>\n",
       "      <td>2018-04-01 12:06:15+00:00</td>\n",
       "      <td>[30.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>268</th>\n",
       "      <td>3</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Ruta 78 / Autopista del Sol</td>\n",
       "      <td>2018-04-01 14:10:44+00:00</td>\n",
       "      <td>[13.0, 13.0, 13.0, 13.0]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>278</th>\n",
       "      <td>3</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Av. Camilo Henríquez</td>\n",
       "      <td>2018-04-01 14:23:36+00:00</td>\n",
       "      <td>[4.0]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     alert_type_id  alert_subtype_id                        street  \\\n",
       "60               3               4.0        Antonia López de Bello   \n",
       "78               3               5.0        Antonia López de Bello   \n",
       "177              3               5.0  Boulevard Lyon Juana de Arco   \n",
       "268              3               4.0   Ruta 78 / Autopista del Sol   \n",
       "278              3               4.0          Av. Camilo Henríquez   \n",
       "\n",
       "              publish_date_dt                       seq  \n",
       "60  2018-04-01 04:07:34+00:00                     [4.0]  \n",
       "78  2018-04-01 04:56:23+00:00                     [4.0]  \n",
       "177 2018-04-01 12:06:15+00:00                    [30.0]  \n",
       "268 2018-04-01 14:10:44+00:00  [13.0, 13.0, 13.0, 13.0]  \n",
       "278 2018-04-01 14:23:36+00:00                     [4.0]  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df[(alerts_df['alert_type_id'] == 3) &\n",
    "          ((alerts_df['alert_subtype_id'] == 4) | (alerts_df['alert_subtype_id'] == 5)) & \n",
    "          (alerts_df.astype(str)['seq'] != '[]')].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50779, 5)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df[(alerts_df['alert_type_id'] == 3) &\n",
    "          ((alerts_df['alert_subtype_id'] == 4) | (alerts_df['alert_subtype_id'] == 5)) & \n",
    "          (alerts_df.astype(str)['seq'] != '[]')].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "alerts_df_w_seq['label'] = np.where((alerts_df_w_seq['alert_type_id'] == 3) &\n",
    "                                    ((alerts_df_w_seq['alert_subtype_id'] == 4) | (alerts_df_w_seq['alert_subtype_id'] == 5)) &\n",
    "                                    (alerts_df_w_seq.astype(str)['seq'] != '[]'), 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "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>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "      <th>seq</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>115494</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:16+00:00</td>\n",
       "      <td>[33.0]</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115508</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:17+00:00</td>\n",
       "      <td>[33.0, 33.0]</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115495</th>\n",
       "      <td>7</td>\n",
       "      <td>33.0</td>\n",
       "      <td>Av. O'Higgins</td>\n",
       "      <td>2018-02-22 18:55:18+00:00</td>\n",
       "      <td>[33.0, 33.0, 33.0]</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        alert_type_id  alert_subtype_id         street  \\\n",
       "115494              7              33.0  Av. O'Higgins   \n",
       "115508              7              33.0  Av. O'Higgins   \n",
       "115495              7              33.0  Av. O'Higgins   \n",
       "\n",
       "                 publish_date_dt                 seq  label  \n",
       "115494 2018-02-22 18:55:16+00:00              [33.0]      0  \n",
       "115508 2018-02-22 18:55:17+00:00        [33.0, 33.0]      0  \n",
       "115495 2018-02-22 18:55:18+00:00  [33.0, 33.0, 33.0]      0  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq[alerts_df_w_seq['label'] == 0].head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(21749, 6)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq[alerts_df_w_seq['label'] == 0].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>alert_type_id</th>\n",
       "      <th>alert_subtype_id</th>\n",
       "      <th>street</th>\n",
       "      <th>publish_date_dt</th>\n",
       "      <th>seq</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>60</th>\n",
       "      <td>3</td>\n",
       "      <td>4.0</td>\n",
       "      <td>Antonia López de Bello</td>\n",
       "      <td>2018-04-01 04:07:34+00:00</td>\n",
       "      <td>[4.0]</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>78</th>\n",
       "      <td>3</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Antonia López de Bello</td>\n",
       "      <td>2018-04-01 04:56:23+00:00</td>\n",
       "      <td>[4.0]</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>177</th>\n",
       "      <td>3</td>\n",
       "      <td>5.0</td>\n",
       "      <td>Boulevard Lyon Juana de Arco</td>\n",
       "      <td>2018-04-01 12:06:15+00:00</td>\n",
       "      <td>[30.0]</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     alert_type_id  alert_subtype_id                        street  \\\n",
       "60               3               4.0        Antonia López de Bello   \n",
       "78               3               5.0        Antonia López de Bello   \n",
       "177              3               5.0  Boulevard Lyon Juana de Arco   \n",
       "\n",
       "              publish_date_dt     seq  label  \n",
       "60  2018-04-01 04:07:34+00:00   [4.0]      1  \n",
       "78  2018-04-01 04:56:23+00:00   [4.0]      1  \n",
       "177 2018-04-01 12:06:15+00:00  [30.0]      1  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq[alerts_df_w_seq['label'] == 1].head(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50779, 6)"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "alerts_df_w_seq[alerts_df_w_seq['label'] == 1].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "91217"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max_len_event_id = alerts_df_w_seq.seq.apply(len).idxmax()\n",
    "max_len_event_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1.0, 4.0, 4.0, 4.0, 4.0, 5.0, 4.0, 4.0, 5.0, 4.0, 4.0, 5.0, 5.0, 30.0, 1.0, 30.0, 3.0, 3.0, 5.0, 30.0, 5.0, 4.0, 5.0, 4.0, 3.0, 5.0, 5.0, 3.0, 4.0, 4.0, 4.0, 5.0, 5.0, 5.0, 5.0, 4.0, 30.0, 5.0, 5.0, 5.0, 4.0, 4.0, 4.0, 5.0, 3.0, 4.0, 5.0, 5.0, 5.0, 5.0, 5.0, 4.0, 4.0, 5.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 5.0, 4.0, 4.0, 4.0, 3.0, 5.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 4.0, 5.0, 5.0, 13.0, 4.0, 5.0, 4.0, 5.0, 5.0, 13.0, 5.0, 4.0, 4.0, 4.0, 4.0, 4.0]\n"
     ]
    }
   ],
   "source": [
    "max_len_event = alerts_df_w_seq.loc[max_len_event_id]\n",
    "print(max_len_event.seq)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "90"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "maxlen = len(max_len_event.seq)\n",
    "maxlen"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets pre process the data in order for keras to understand it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:527: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:528: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:529: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:530: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/dtypes.py:535: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n"
     ]
    }
   ],
   "source": [
    "#from keras.utils import to_categorical\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[[33.0], [33.0, 33.0], [33.0, 33.0, 33.0], [33.0], [3.0]]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sequences = alerts_df_w_seq.seq.tolist()\n",
    "sequences[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0,  0,  0, ...,  0,  0, 33],\n",
       "       [ 0,  0,  0, ...,  0, 33, 33],\n",
       "       [ 0,  0,  0, ..., 33, 33, 33],\n",
       "       ...,\n",
       "       [ 0,  0,  0, ...,  0,  0, 29],\n",
       "       [ 0,  0,  0, ...,  0, 29, 29],\n",
       "       [ 0,  0,  0, ..., 29, 29, 10]], dtype=int32)"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pad_sequences(sequences, maxlen=maxlen)\n",
    "data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = np.array(alerts_df_w_seq.label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(12)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = np.arange(data.shape[0])\n",
    "np.random.shuffle(indices)\n",
    "data = data[indices]\n",
    "labels = labels[indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_samples = int(len(indices) * .5)\n",
    "validation_samples = len(indices) - training_samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = data[:training_samples]\n",
    "y_train = labels[:training_samples]\n",
    "X_valid = data[training_samples: training_samples + validation_samples]\n",
    "y_valid = labels[training_samples: training_samples + validation_samples]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0,  0,  0, ..., 30,  1,  3],\n",
       "       [ 0,  0,  0, ...,  0,  0,  4],\n",
       "       [ 0,  0,  0, ...,  4,  5,  4],\n",
       "       ...,\n",
       "       [ 0,  0,  0, ...,  0,  4, 29],\n",
       "       [ 0,  0,  0, ...,  3, 30,  5],\n",
       "       [ 0,  0,  0, ...,  4,  5,  4]], dtype=int32)"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From here we set up our neuronal networks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "35"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_events = 34 + 1\n",
    "num_events"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_dim = 20"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "embedding_matrix = np.random.rand(num_events, embedding_dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, Flatten, Dense, LSTM\n",
    "\n",
    "units = 32 # more implies more complexity, unrelated with\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(num_events, embedding_dim))\n",
    "model.add(LSTM(units))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.layers[0].set_weights([embedding_matrix])\n",
    "model.layers[0].trainable = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_1 (Embedding)      (None, None, 20)          700       \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 32)                6784      \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 7,517\n",
      "Trainable params: 6,817\n",
      "Non-trainable params: 700\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Train on 36264 samples, validate on 36264 samples\n",
      "Epoch 1/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5670 - acc: 0.7218 - val_loss: 0.5592 - val_acc: 0.7337\n",
      "Epoch 2/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5521 - acc: 0.7350 - val_loss: 0.5648 - val_acc: 0.7290\n",
      "Epoch 3/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5491 - acc: 0.7374 - val_loss: 0.5518 - val_acc: 0.7394\n",
      "Epoch 4/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5482 - acc: 0.7385 - val_loss: 0.5550 - val_acc: 0.7349\n",
      "Epoch 5/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5470 - acc: 0.7408 - val_loss: 0.5498 - val_acc: 0.7396\n",
      "Epoch 6/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5463 - acc: 0.7405 - val_loss: 0.5496 - val_acc: 0.7390\n",
      "Epoch 7/20\n",
      "36264/36264 [==============================] - 98s 3ms/step - loss: 0.5456 - acc: 0.7403 - val_loss: 0.5528 - val_acc: 0.7369\n",
      "Epoch 8/20\n",
      "36264/36264 [==============================] - 98s 3ms/step - loss: 0.5455 - acc: 0.7407 - val_loss: 0.5557 - val_acc: 0.7323\n",
      "Epoch 9/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5452 - acc: 0.7425 - val_loss: 0.5487 - val_acc: 0.7402\n",
      "Epoch 10/20\n",
      "36264/36264 [==============================] - 98s 3ms/step - loss: 0.5456 - acc: 0.7396 - val_loss: 0.5490 - val_acc: 0.7392\n",
      "Epoch 11/20\n",
      "36264/36264 [==============================] - 98s 3ms/step - loss: 0.5450 - acc: 0.7423 - val_loss: 0.5490 - val_acc: 0.7385\n",
      "Epoch 12/20\n",
      "36264/36264 [==============================] - 98s 3ms/step - loss: 0.5446 - acc: 0.7426 - val_loss: 0.5515 - val_acc: 0.7386\n",
      "Epoch 13/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5442 - acc: 0.7420 - val_loss: 0.5494 - val_acc: 0.7389\n",
      "Epoch 14/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5439 - acc: 0.7405 - val_loss: 0.5491 - val_acc: 0.7399\n",
      "Epoch 15/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5440 - acc: 0.7415 - val_loss: 0.5507 - val_acc: 0.7387\n",
      "Epoch 16/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5439 - acc: 0.7425 - val_loss: 0.5493 - val_acc: 0.7394\n",
      "Epoch 17/20\n",
      "36264/36264 [==============================] - 97s 3ms/step - loss: 0.5435 - acc: 0.7417 - val_loss: 0.5489 - val_acc: 0.7404\n",
      "Epoch 18/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5435 - acc: 0.7419 - val_loss: 0.5497 - val_acc: 0.7405\n",
      "Epoch 19/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5434 - acc: 0.7423 - val_loss: 0.5531 - val_acc: 0.7366\n",
      "Epoch 20/20\n",
      "36264/36264 [==============================] - 96s 3ms/step - loss: 0.5431 - acc: 0.7427 - val_loss: 0.5489 - val_acc: 0.7407\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(X_train, y_train,\n",
    "                    epochs=20,\n",
    "                    batch_size=32,\n",
    "                    validation_data=(X_valid, y_valid))\n",
    "model.save(\"mymodel_embedding_untrainable.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "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"
    },
    {
     "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "So next step, we can do an experiment to allow the Embedding layer be trained and adjusted."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, Flatten, Dense, LSTM\n",
    "\n",
    "units = 32\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(num_events, embedding_dim))\n",
    "model.add(LSTM(units))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.layers[0].set_weights([embedding_matrix])\n",
    "model.layers[0].trainable = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 36264 samples, validate on 36264 samples\n",
      "Epoch 1/20\n",
      "36264/36264 [==============================] - 103s 3ms/step - loss: 0.5601 - acc: 0.7288 - val_loss: 0.5590 - val_acc: 0.7293\n",
      "Epoch 2/20\n",
      "36264/36264 [==============================] - 103s 3ms/step - loss: 0.5476 - acc: 0.7400 - val_loss: 0.5504 - val_acc: 0.7378\n",
      "Epoch 3/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5459 - acc: 0.7417 - val_loss: 0.5492 - val_acc: 0.7394\n",
      "Epoch 4/20\n",
      "36264/36264 [==============================] - 103s 3ms/step - loss: 0.5449 - acc: 0.7420 - val_loss: 0.5505 - val_acc: 0.7386\n",
      "Epoch 5/20\n",
      "36264/36264 [==============================] - 104s 3ms/step - loss: 0.5444 - acc: 0.7409 - val_loss: 0.5484 - val_acc: 0.7395\n",
      "Epoch 6/20\n",
      "36264/36264 [==============================] - 103s 3ms/step - loss: 0.5438 - acc: 0.7419 - val_loss: 0.5503 - val_acc: 0.7383\n",
      "Epoch 7/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5437 - acc: 0.7421 - val_loss: 0.5521 - val_acc: 0.7388\n",
      "Epoch 8/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5432 - acc: 0.7430 - val_loss: 0.5499 - val_acc: 0.7388\n",
      "Epoch 9/20\n",
      "36264/36264 [==============================] - 103s 3ms/step - loss: 0.5427 - acc: 0.7432 - val_loss: 0.5500 - val_acc: 0.7394\n",
      "Epoch 10/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5426 - acc: 0.7430 - val_loss: 0.5485 - val_acc: 0.7394\n",
      "Epoch 11/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5423 - acc: 0.7433 - val_loss: 0.5513 - val_acc: 0.7376\n",
      "Epoch 12/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5421 - acc: 0.7435 - val_loss: 0.5495 - val_acc: 0.7395\n",
      "Epoch 13/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5418 - acc: 0.7430 - val_loss: 0.5489 - val_acc: 0.7389\n",
      "Epoch 14/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5419 - acc: 0.7440 - val_loss: 0.5486 - val_acc: 0.7397\n",
      "Epoch 15/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5415 - acc: 0.7434 - val_loss: 0.5491 - val_acc: 0.7387\n",
      "Epoch 16/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5414 - acc: 0.7437 - val_loss: 0.5493 - val_acc: 0.7388\n",
      "Epoch 17/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5413 - acc: 0.7441 - val_loss: 0.5504 - val_acc: 0.7392\n",
      "Epoch 18/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5409 - acc: 0.7443 - val_loss: 0.5490 - val_acc: 0.7398\n",
      "Epoch 19/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5408 - acc: 0.7445 - val_loss: 0.5495 - val_acc: 0.7392\n",
      "Epoch 20/20\n",
      "36264/36264 [==============================] - 102s 3ms/step - loss: 0.5407 - acc: 0.7440 - val_loss: 0.5491 - val_acc: 0.7402\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(X_train, y_train,\n",
    "                    epochs=20,\n",
    "                    batch_size=32,\n",
    "                    validation_data=(X_valid, y_valid))\n",
    "model.save(\"mymodel_embedding_trainable.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "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"
    },
    {
     "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will add two parameters related with Dropouts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /opt/conda/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Embedding, Flatten, Dense, LSTM\n",
    "\n",
    "units = 32\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(num_events, embedding_dim))\n",
    "model.add(LSTM(units, dropout=0.2, recurrent_dropout=0.2))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.layers[0].set_weights([embedding_matrix])\n",
    "model.layers[0].trainable = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 36264 samples, validate on 36264 samples\n",
      "Epoch 1/20\n",
      "36264/36264 [==============================] - 126s 3ms/step - loss: 0.5691 - acc: 0.7197 - val_loss: 0.5630 - val_acc: 0.7283\n",
      "Epoch 2/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5513 - acc: 0.7377 - val_loss: 0.5504 - val_acc: 0.7388\n",
      "Epoch 3/20\n",
      "36264/36264 [==============================] - 124s 3ms/step - loss: 0.5486 - acc: 0.7393 - val_loss: 0.5493 - val_acc: 0.7395\n",
      "Epoch 4/20\n",
      "36264/36264 [==============================] - 124s 3ms/step - loss: 0.5462 - acc: 0.7405 - val_loss: 0.5540 - val_acc: 0.7343\n",
      "Epoch 5/20\n",
      "36264/36264 [==============================] - 122s 3ms/step - loss: 0.5460 - acc: 0.7400 - val_loss: 0.5538 - val_acc: 0.7331\n",
      "Epoch 6/20\n",
      "36264/36264 [==============================] - 127s 3ms/step - loss: 0.5452 - acc: 0.7412 - val_loss: 0.5487 - val_acc: 0.7385\n",
      "Epoch 7/20\n",
      "36264/36264 [==============================] - 126s 3ms/step - loss: 0.5447 - acc: 0.7416 - val_loss: 0.5496 - val_acc: 0.7340\n",
      "Epoch 8/20\n",
      "36264/36264 [==============================] - 127s 4ms/step - loss: 0.5441 - acc: 0.7420 - val_loss: 0.5490 - val_acc: 0.7390\n",
      "Epoch 9/20\n",
      "36264/36264 [==============================] - 126s 3ms/step - loss: 0.5439 - acc: 0.7423 - val_loss: 0.5519 - val_acc: 0.7396\n",
      "Epoch 10/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5437 - acc: 0.7420 - val_loss: 0.5483 - val_acc: 0.7394\n",
      "Epoch 11/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5435 - acc: 0.7421 - val_loss: 0.5483 - val_acc: 0.7395\n",
      "Epoch 12/20\n",
      "36264/36264 [==============================] - 127s 3ms/step - loss: 0.5439 - acc: 0.7428 - val_loss: 0.5489 - val_acc: 0.7400\n",
      "Epoch 13/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5434 - acc: 0.7429 - val_loss: 0.5486 - val_acc: 0.7394\n",
      "Epoch 14/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5430 - acc: 0.7426 - val_loss: 0.5502 - val_acc: 0.7384\n",
      "Epoch 15/20\n",
      "36264/36264 [==============================] - 124s 3ms/step - loss: 0.5430 - acc: 0.7429 - val_loss: 0.5494 - val_acc: 0.7399\n",
      "Epoch 16/20\n",
      "36264/36264 [==============================] - 126s 3ms/step - loss: 0.5431 - acc: 0.7422 - val_loss: 0.5484 - val_acc: 0.7401\n",
      "Epoch 17/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5427 - acc: 0.7420 - val_loss: 0.5486 - val_acc: 0.7397\n",
      "Epoch 18/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5429 - acc: 0.7424 - val_loss: 0.5485 - val_acc: 0.7399\n",
      "Epoch 19/20\n",
      "36264/36264 [==============================] - 124s 3ms/step - loss: 0.5427 - acc: 0.7434 - val_loss: 0.5493 - val_acc: 0.7402\n",
      "Epoch 20/20\n",
      "36264/36264 [==============================] - 125s 3ms/step - loss: 0.5423 - acc: 0.7428 - val_loss: 0.5508 - val_acc: 0.7400\n"
     ]
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['acc'])\n",
    "history = model.fit(X_train, y_train,\n",
    "                    epochs=20,\n",
    "                    batch_size=32,\n",
    "                    validation_data=(X_valid, y_valid))\n",
    "model.save(\"mymodel_embedding_trainable_with_dropout.h5\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "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"
    },
    {
     "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets analyze our results more in depth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_valid_hat =  model.predict_classes(X_valid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0],\n",
       "       [1],\n",
       "       [1],\n",
       "       ...,\n",
       "       [1],\n",
       "       [1],\n",
       "       [1]], dtype=int32)"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_valid_hat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "tn, fp, fn, tp = confusion_matrix(y_valid, y_valid_hat).ravel()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2994, 7956, 1471, 23843)"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(tn, fp, fn, tp)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}