service-usage-machine-learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BiciMAD - Machine learning predictions\n",
    "\n",
    "Copyright © 2017 Javi Ramírez <javi.rmrz@gmail.com> | [@rameerez [tw]](http://twitter.com/rameerez) | [GitHub](http://github.com/rameerez)\n",
    "\n",
    "This code is Open Source, released under the MIT License."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Why/what\n",
    "We've studied the actual BiciMAD users' behavior in other notebooks – but can we predict user behavior using Artificial Intelligence?\n",
    "![crowd](img/crowd.jpg)\n",
    "[Photo credit](https://unsplash.com/photos/ABGaVhJxwDQ)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data sources\n",
    "\n",
    "Apart from BiciMAD's data, we're using meteo data. All credits to [AEMET OpenData](http://www.aemet.es/es/datos_abiertos/AEMET_OpenData) (Spanish Meteorological Agency).\n",
    "\n",
    "I tried to get **hourly** meteo observations for the month of March 2017, but couldn't find any source. Therefore we'll be grouping data by day."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Definitions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Data describing all bike rides\n",
    "BIKE_RIDES_DATASET = './data/export_20170301_20170406.json'\n",
    "# Data describing all stations (used just for getting stations' coordinates & names)\n",
    "STATIONS_DATASET = './data/stations.json'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# AEMET's (Spanish Meteorological Agency) observations in Madrid (Retiro) for each day in March 2017\n",
    "MARCH_WEATHER_DATASET = './data/march_weather_madrid_retiro.json'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib as plt\n",
    "import math\n",
    "import dateutil.parser\n",
    "from pandas.io.json import json_normalize\n",
    "import tflearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df = pd.read_json(BIKE_RIDES_DATASET, lines=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>_id</th>\n",
       "      <th>ageRange</th>\n",
       "      <th>idplug_base</th>\n",
       "      <th>idplug_station</th>\n",
       "      <th>idunplug_base</th>\n",
       "      <th>idunplug_station</th>\n",
       "      <th>track</th>\n",
       "      <th>travel_time</th>\n",
       "      <th>unplug_hourTime</th>\n",
       "      <th>user_day_code</th>\n",
       "      <th>user_type</th>\n",
       "      <th>zip_code</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>{'$oid': '58d1b3dd865e9c30d0b83c6f'}</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>17</td>\n",
       "      <td>5</td>\n",
       "      <td>14</td>\n",
       "      <td>NaN</td>\n",
       "      <td>2303</td>\n",
       "      <td>{'$date': '2017-03-01T00:00:00.000+0100'}</td>\n",
       "      <td>cc6d426ada96494d8dfe0fe0af512599f439cd9413d50c...</td>\n",
       "      <td>1</td>\n",
       "      <td>28012</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>{'$oid': '58d1b3dc865e9c30d0b83c6c'}</td>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>76</td>\n",
       "      <td>22</td>\n",
       "      <td>123</td>\n",
       "      <td>NaN</td>\n",
       "      <td>876</td>\n",
       "      <td>{'$date': '2017-03-01T00:00:00.000+0100'}</td>\n",
       "      <td>a81b44c3fe3c2f1e4bf723d228a0d025962228179c2e4d...</td>\n",
       "      <td>1</td>\n",
       "      <td>28009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>{'$oid': '58d1b3dc865e9c30d0b83c6a'}</td>\n",
       "      <td>5</td>\n",
       "      <td>16</td>\n",
       "      <td>163</td>\n",
       "      <td>19</td>\n",
       "      <td>166</td>\n",
       "      <td>NaN</td>\n",
       "      <td>483</td>\n",
       "      <td>{'$date': '2017-03-01T00:00:00.000+0100'}</td>\n",
       "      <td>b20f0b657c493842b0aa5ebd32eed5d940c699715ee8bc...</td>\n",
       "      <td>1</td>\n",
       "      <td>28010</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                    _id  ageRange  idplug_base  \\\n",
       "0  {'$oid': '58d1b3dd865e9c30d0b83c6f'}         4            4   \n",
       "1  {'$oid': '58d1b3dc865e9c30d0b83c6c'}         4            2   \n",
       "2  {'$oid': '58d1b3dc865e9c30d0b83c6a'}         5           16   \n",
       "\n",
       "   idplug_station  idunplug_base  idunplug_station track  travel_time  \\\n",
       "0              17              5                14   NaN         2303   \n",
       "1              76             22               123   NaN          876   \n",
       "2             163             19               166   NaN          483   \n",
       "\n",
       "                             unplug_hourTime  \\\n",
       "0  {'$date': '2017-03-01T00:00:00.000+0100'}   \n",
       "1  {'$date': '2017-03-01T00:00:00.000+0100'}   \n",
       "2  {'$date': '2017-03-01T00:00:00.000+0100'}   \n",
       "\n",
       "                                       user_day_code  user_type zip_code  \n",
       "0  cc6d426ada96494d8dfe0fe0af512599f439cd9413d50c...          1    28012  \n",
       "1  a81b44c3fe3c2f1e4bf723d228a0d025962228179c2e4d...          1    28009  \n",
       "2  b20f0b657c493842b0aa5ebd32eed5d940c699715ee8bc...          1    28010  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[:3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Converting unplug time JSON to valid Datetime\n",
    "Unfortunately the `unplug_hourTime` comes in an unnecesarily descriptive JSON. Let's convert that to a valid `Datetime` so we can do stuff like `date.month`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def parse_unplug_datetime(row):\n",
    "    date_as_string = row['unplug_hourTime']['$date']\n",
    "    return dateutil.parser.parse(date_as_string)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df['unplug_datetime'] = df.apply(parse_unplug_datetime, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Yay! Valid datetimes. Let's now group by month and day and plot the number of rides by day:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "times = pd.DatetimeIndex(df.unplug_datetime)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x21662d978>"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEBCAYAAACT92m7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x216640940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.groupby([times.month, times.day])._id.count().plot()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset contains data for the entire month of March and some days of April. Let's filter and get just March data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "march_df = df[times.month == 3]\n",
    "march_times = pd.DatetimeIndex(march_df.unplug_datetime)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x21665d198>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEiCAYAAAASzx4jAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2165827b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "march_df.groupby([march_times.month, march_times.day])._id.count().plot(legend=False, rot=90, xticks=pd.Series(range(1,31)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Does weather affect bike usage?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Caution! orient='records' is required!\n",
    "march_weather_df = pd.read_json(MARCH_WEATHER_DATASET, orient='records')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "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>altitud</th>\n",
       "      <th>dir</th>\n",
       "      <th>fecha</th>\n",
       "      <th>horaPresMax</th>\n",
       "      <th>horaPresMin</th>\n",
       "      <th>horaracha</th>\n",
       "      <th>horatmax</th>\n",
       "      <th>horatmin</th>\n",
       "      <th>indicativo</th>\n",
       "      <th>nombre</th>\n",
       "      <th>prec</th>\n",
       "      <th>presMax</th>\n",
       "      <th>presMin</th>\n",
       "      <th>provincia</th>\n",
       "      <th>racha</th>\n",
       "      <th>tmax</th>\n",
       "      <th>tmed</th>\n",
       "      <th>tmin</th>\n",
       "      <th>velmedia</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>667</td>\n",
       "      <td>26</td>\n",
       "      <td>2017-03-01</td>\n",
       "      <td>11</td>\n",
       "      <td>00</td>\n",
       "      <td>00:00</td>\n",
       "      <td>16:10</td>\n",
       "      <td>06:50</td>\n",
       "      <td>3195</td>\n",
       "      <td>MADRID, RETIRO</td>\n",
       "      <td>0,0</td>\n",
       "      <td>944,2</td>\n",
       "      <td>940,5</td>\n",
       "      <td>MADRID</td>\n",
       "      <td>6,9</td>\n",
       "      <td>16,0</td>\n",
       "      <td>10,6</td>\n",
       "      <td>5,2</td>\n",
       "      <td>1,1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   altitud  dir       fecha horaPresMax horaPresMin horaracha horatmax  \\\n",
       "0      667   26  2017-03-01          11          00     00:00    16:10   \n",
       "\n",
       "  horatmin  indicativo          nombre prec presMax presMin provincia racha  \\\n",
       "0    06:50        3195  MADRID, RETIRO  0,0   944,2   940,5    MADRID   6,9   \n",
       "\n",
       "   tmax  tmed tmin velmedia  \n",
       "0  16,0  10,6  5,2      1,1  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "march_weather_df[:1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well, shit. Is nobody going to record their data in a freaking valid way? Those \"float\" numbers are NOT numbers – they are strings! The only valid float separator is a point `.`, not a comma `,`.\n",
    "\n",
    "And that date... that is a string, too. Gosh.\n",
    "\n",
    "Time to clean (again) our data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def comma_string_to_float(float_string):\n",
    "    return float(float_string.replace(\".\",\"\").replace(\",\",\".\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def operations(row):\n",
    "  df1 = comma_string_to_float(row['prec'])\n",
    "  df2 = comma_string_to_float(row['tmed'])\n",
    "  return df1, df2\n",
    "\n",
    "march_weather_df[\"rain\"], march_weather_df[\"avgtemp\"] = zip(*march_weather_df.apply(lambda row: operations(row), axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def fecha_to_date(row):\n",
    "    return dateutil.parser.parse(row['fecha'])\n",
    "\n",
    "march_weather_df['datetime'] = march_weather_df.apply(fecha_to_date, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "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>datetime</th>\n",
       "      <th>rain</th>\n",
       "      <th>avgtemp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2017-03-01</td>\n",
       "      <td>0.0</td>\n",
       "      <td>10.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2017-03-02</td>\n",
       "      <td>0.0</td>\n",
       "      <td>11.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2017-03-03</td>\n",
       "      <td>2.1</td>\n",
       "      <td>8.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    datetime  rain  avgtemp\n",
       "0 2017-03-01   0.0     10.6\n",
       "1 2017-03-02   0.0     11.0\n",
       "2 2017-03-03   2.1      8.0"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "march_weather_df[['datetime','rain','avgtemp']][:3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And that would be how you get valid float numbers for meteo observations. Let's plot rain:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x215d61be0>"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW4AAAD/CAYAAADR7zzdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2160edf28>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "march_weather_df.rain.plot(rot=90, xticks=pd.Series(range(1,31)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Interesting! Rain seems to be the inverse of usage! Let's recall:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x154087128>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYcAAAEiCAYAAAASzx4jAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x154029828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "march_df.groupby([march_times.month, march_times.day])._id.count().plot(legend=False, rot=90, xticks=pd.Series(range(1,31)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, the days with the most rain are the days with the least bike rides."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## From daily to hourly meteo data\n",
    "\n",
    "So, AEMET won't provide hourly observations. Instead, they provide daily observations.\n",
    "\n",
    "Ideas:\n",
    "\n",
    "- **Temp**: mintemp & maxtemp are available, along with the time at which both ocurred. Avgtemp is available, too. Strategy idea: linear distribution from 00:00 to mintemp time (last day temp to mintemp), then from mintemp time to maxtemp time and so on.\n",
    "- **Pressure**: same strategy as temperature.\n",
    "- **Rain**: the most naive assumption would be hourly precipitation = daily/24, but we can add some precision taking into account the previous and next day's precipitation to make it an ascending, descending or normal distribution.\n",
    "- **Wind speed**: Only avg is provided. For simplicity we could assume the hourly wind speed to be just the average for the day."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Machine learning\n",
    "\n",
    "Let's go ahead and use Artificial Intelligence to train a neural network to predict number of bike trips per day depending on rain and average temperature for that day."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data preparation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "First, let's create a dataframe with the usage per day."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "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>unplug_datetime</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2017-03-01 00:00:00+01:00</td>\n",
       "      <td>10420</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2017-03-02 00:00:00+01:00</td>\n",
       "      <td>10174</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2017-03-03 00:00:00+01:00</td>\n",
       "      <td>6762</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            unplug_datetime  count\n",
       "0 2017-03-01 00:00:00+01:00  10420\n",
       "1 2017-03-02 00:00:00+01:00  10174\n",
       "2 2017-03-03 00:00:00+01:00   6762"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_usage_by_day = march_df.set_index('unplug_datetime').groupby(pd.Grouper(freq='D'))._id.count().to_frame(name=\"count\").reset_index()\n",
    "df_usage_by_day[:3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then, we'll need to join it with our weather data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "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>datetime</th>\n",
       "      <th>rain</th>\n",
       "      <th>avgtemp</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2017-03-01</td>\n",
       "      <td>0.0</td>\n",
       "      <td>10.6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2017-03-02</td>\n",
       "      <td>0.0</td>\n",
       "      <td>11.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2017-03-03</td>\n",
       "      <td>2.1</td>\n",
       "      <td>8.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    datetime  rain  avgtemp\n",
       "0 2017-03-01   0.0     10.6\n",
       "1 2017-03-02   0.0     11.0\n",
       "2 2017-03-03   2.1      8.0"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "march_weather_df[['datetime','rain','avgtemp']][:3]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well, more stuff to clean. Seems like our BiciMAD data now has `datetime.datetime` timestamps and our AEMET data has `datetime.date` timestamps. Let's remove the hour, minutes and seconds and make it timezone naive."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df_usage_by_day['unplug_date'] = df_usage_by_day['unplug_datetime'].dt.date"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "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>unplug_datetime</th>\n",
       "      <th>count</th>\n",
       "      <th>unplug_date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2017-03-01 00:00:00+01:00</td>\n",
       "      <td>10420</td>\n",
       "      <td>2017-03-01</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            unplug_datetime  count unplug_date\n",
       "0 2017-03-01 00:00:00+01:00  10420  2017-03-01"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_usage_by_day[:1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Yay! Let's join the tables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "daily_frequencies_with_meteo_df = march_weather_df.set_index('datetime').join(df_usage_by_day.set_index('unplug_date')).reset_index()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "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>datetime</th>\n",
       "      <th>rain</th>\n",
       "      <th>avgtemp</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2017-03-01</td>\n",
       "      <td>0.0</td>\n",
       "      <td>10.6</td>\n",
       "      <td>10420.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    datetime  rain  avgtemp    count\n",
       "0 2017-03-01   0.0     10.6  10420.0"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "daily_frequencies_with_meteo_df[['datetime','rain','avgtemp','count']][:1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well and we kind of need to know the day of the week instead of the actual date:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "daily_frequencies_with_meteo_df['dayofweek'] = daily_frequencies_with_meteo_df['datetime'].dt.dayofweek"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "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>dayofweek</th>\n",
       "      <th>rain</th>\n",
       "      <th>avgtemp</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>10.6</td>\n",
       "      <td>10420.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   dayofweek  rain  avgtemp    count\n",
       "0          2   0.0     10.6  10420.0"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "daily_frequencies_with_meteo_df[['dayofweek','rain','avgtemp','count']][:1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note: The day of the week with Monday=0, Sunday=6"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Let's use a deep neural network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "neural_net_data = daily_frequencies_with_meteo_df[['dayofweek','rain','avgtemp','count']].copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just for convenience, let's save the dataframe we'll use in a separate .csv file. This way, if the training goes wrong and we have to restart the kernel, we won't have to rerun all dataprep operations again."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "neural_net_data.to_csv('data/neural_net_data.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And let's import that just to make sure we're using the .csv data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "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>dayofweek</th>\n",
       "      <th>rain</th>\n",
       "      <th>avgtemp</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2</td>\n",
       "      <td>0.0</td>\n",
       "      <td>10.6</td>\n",
       "      <td>10420.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   dayofweek  rain  avgtemp    count\n",
       "0          2   0.0     10.6  10420.0"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neural_net_data = pd.read_csv('data/neural_net_data.csv', index_col=0)\n",
    "neural_net_data[:1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now over to the TFLearn part. Note that there is no need to read the .csv prior to this point, we're just doing it for illustration purposes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Load CSV file, indicate that the fifth (3) column represents labels. Ignore the first column (it just contains IDs)\n",
    "from tflearn.data_utils import load_csv\n",
    "data, labels = load_csv('data/neural_net_data.csv', target_column=4, columns_to_ignore=[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This creates two variables.\n",
    "\n",
    "These, as described by the [TFLearn documentation](http://tflearn.org/models/dnn/) are the two datasets required to train our Deep Neural Network.\n",
    "\n",
    "Please note that **a neural network model can only perform operations over numbers**. We need to convert every non-numerical label/value to a numerical representation before this point. It's not the case, so we continue.\n",
    "\n",
    "The first variable, `data`, our `X_inputs` a list, array or ndarray. Data to feed to train model.\n",
    "The second one, `labels`, our targets (Labels) to feed to train model.\n",
    "\n",
    "---\n",
    "\n",
    "We're now going to create the Deep Neural Network. We declare its shape as `[None, 3]`, meaning that we have unknown rows (`None` dimension) and that our data has `3` features (`dayofweek`,`rain`,`avgtemp`)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Build neural network\n",
    "net = tflearn.input_data(shape=[None, 3])\n",
    "net = tflearn.fully_connected(net, 32)\n",
    "net = tflearn.fully_connected(net, 1)\n",
    "net = tflearn.regression(net, optimizer='sgd', loss='mean_square', learning_rate=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Need to do this so we don't get dimension error when training the neural net\n",
    "labels = np.asarray(labels)[:, np.newaxis]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Step: 199  | total loss: \u001b[1m\u001b[32m3195796.00000\u001b[0m\u001b[0m | time: 0.003s\n",
      "| SGD | epoch: 100 | loss: 3195796.00000 - binary_acc: 0.0000 -- iter: 16/31\n",
      "Training Step: 200  | total loss: \u001b[1m\u001b[32m3325802.00000\u001b[0m\u001b[0m | time: 0.006s\n",
      "| SGD | epoch: 100 | loss: 3325802.00000 - binary_acc: 0.0000 -- iter: 31/31\n",
      "--\n"
     ]
    }
   ],
   "source": [
    "# Define model\n",
    "model = tflearn.DNN(net)\n",
    "# Start training (apply gradient descent algorithm)\n",
    "model.fit(data, labels, n_epoch=100, batch_size=16, show_metric=True, snapshot_epoch=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Neural net trained! Let's test it with an example day.\n",
    "\n",
    "Say day 10 of our data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dayofweek       5.0\n",
       "rain            0.0\n",
       "avgtemp        16.1\n",
       "count        9402.0\n",
       "Name: 10, dtype: float64"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neural_net_data.iloc[10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Weekday 5 (Saturday), rain 0.0, avgtemp 16.1. Had a total of 9402 travels.\n",
    "\n",
    "Let's test it:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction: 11933.9\n"
     ]
    }
   ],
   "source": [
    "day = [5.0, 0.0, 16.1]\n",
    "pred = model.predict([day])\n",
    "print(\"Prediction:\", pred[0][0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Well, not so bad considering we only have 31 datapoints (which is AWFUL, we need to get moar data and/or find the way to get hourly datapoints)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "\n",
    "## To-do\n",
    "\n",
    " - Get, or transform/distribute AEMET's daily data into hours and get hourly usage data\n",
    " - Use other meteo info to feed the neural net: pressure, wind, etc.\n",
    " - Could we get hourly observations about sun brightness, cloud cover and other meteo data?\n",
    " - Get data from the whole year instead of only one month\n",
    " - Predict usage **by station**\n",
    " - User other external data that might influence service usage, such as public events or national holidays"
   ]
  }
 ],
 "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": "3.5.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}