TP_Geoloc_Msp_Telecom_0.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import math\n",
    "from geopy.distance import vincenty\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# load train and test data\n",
    "df_mess_train = pd.read_csv('mess_train_list.csv') # train set\n",
    "df_mess_test = pd.read_csv('mess_test_list.csv') # test set\n",
    "pos_train = pd.read_csv('pos_train_list.csv') # position associated to train set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>objid</th>\n",
       "      <th>bsid</th>\n",
       "      <th>did</th>\n",
       "      <th>nseq</th>\n",
       "      <th>rssi</th>\n",
       "      <th>time_ux</th>\n",
       "      <th>bs_lat</th>\n",
       "      <th>bs_lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>573bf1d9864fce1a9af8c5c9</td>\n",
       "      <td>2841</td>\n",
       "      <td>473335.0</td>\n",
       "      <td>0.5</td>\n",
       "      <td>-121.5</td>\n",
       "      <td>1.463546e+12</td>\n",
       "      <td>39.617794</td>\n",
       "      <td>-104.954917</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>573bf1d9864fce1a9af8c5c9</td>\n",
       "      <td>3526</td>\n",
       "      <td>473335.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>-125.0</td>\n",
       "      <td>1.463546e+12</td>\n",
       "      <td>39.677251</td>\n",
       "      <td>-104.952721</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>573bf3533e952e19126b256a</td>\n",
       "      <td>2605</td>\n",
       "      <td>473335.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-134.0</td>\n",
       "      <td>1.463547e+12</td>\n",
       "      <td>39.612745</td>\n",
       "      <td>-105.008827</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>573c0cd0f0fe6e735a699b93</td>\n",
       "      <td>2610</td>\n",
       "      <td>473953.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>-132.0</td>\n",
       "      <td>1.463553e+12</td>\n",
       "      <td>39.797969</td>\n",
       "      <td>-105.073460</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>573c0cd0f0fe6e735a699b93</td>\n",
       "      <td>3574</td>\n",
       "      <td>473953.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-120.0</td>\n",
       "      <td>1.463553e+12</td>\n",
       "      <td>39.723151</td>\n",
       "      <td>-104.956216</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      objid  bsid       did  nseq   rssi       time_ux  \\\n",
       "0  573bf1d9864fce1a9af8c5c9  2841  473335.0   0.5 -121.5  1.463546e+12   \n",
       "1  573bf1d9864fce1a9af8c5c9  3526  473335.0   2.0 -125.0  1.463546e+12   \n",
       "2  573bf3533e952e19126b256a  2605  473335.0   1.0 -134.0  1.463547e+12   \n",
       "3  573c0cd0f0fe6e735a699b93  2610  473953.0   2.0 -132.0  1.463553e+12   \n",
       "4  573c0cd0f0fe6e735a699b93  3574  473953.0   1.0 -120.0  1.463553e+12   \n",
       "\n",
       "      bs_lat      bs_lng  \n",
       "0  39.617794 -104.954917  \n",
       "1  39.677251 -104.952721  \n",
       "2  39.612745 -105.008827  \n",
       "3  39.797969 -105.073460  \n",
       "4  39.723151 -104.956216  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_mess_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>bsid</th>\n",
       "      <th>did</th>\n",
       "      <th>nseq</th>\n",
       "      <th>rssi</th>\n",
       "      <th>time_ux</th>\n",
       "      <th>bs_lat</th>\n",
       "      <th>bs_lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>39250.000000</td>\n",
       "      <td>3.925000e+04</td>\n",
       "      <td>39250.000000</td>\n",
       "      <td>39250.000000</td>\n",
       "      <td>3.925000e+04</td>\n",
       "      <td>39250.000000</td>\n",
       "      <td>39250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>4533.774752</td>\n",
       "      <td>4.835932e+05</td>\n",
       "      <td>1.043541</td>\n",
       "      <td>-125.863580</td>\n",
       "      <td>1.473368e+12</td>\n",
       "      <td>42.554998</td>\n",
       "      <td>-100.969685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>2570.495638</td>\n",
       "      <td>9.893140e+04</td>\n",
       "      <td>0.563540</td>\n",
       "      <td>9.226500</td>\n",
       "      <td>5.506910e+09</td>\n",
       "      <td>7.722787</td>\n",
       "      <td>11.522732</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>879.000000</td>\n",
       "      <td>4.732880e+05</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>-155.000000</td>\n",
       "      <td>1.463546e+12</td>\n",
       "      <td>36.053394</td>\n",
       "      <td>-107.463146</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>2790.000000</td>\n",
       "      <td>4.761230e+05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-132.500000</td>\n",
       "      <td>1.468410e+12</td>\n",
       "      <td>39.747448</td>\n",
       "      <td>-105.074287</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3559.000000</td>\n",
       "      <td>4.763150e+05</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-127.333333</td>\n",
       "      <td>1.473444e+12</td>\n",
       "      <td>39.793585</td>\n",
       "      <td>-105.043685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>7456.000000</td>\n",
       "      <td>4.766090e+05</td>\n",
       "      <td>1.500000</td>\n",
       "      <td>-121.000000</td>\n",
       "      <td>1.479118e+12</td>\n",
       "      <td>39.838740</td>\n",
       "      <td>-104.998429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>11951.000000</td>\n",
       "      <td>1.747448e+06</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>-64.000000</td>\n",
       "      <td>1.481143e+12</td>\n",
       "      <td>64.300000</td>\n",
       "      <td>-68.500000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               bsid           did          nseq          rssi       time_ux  \\\n",
       "count  39250.000000  3.925000e+04  39250.000000  39250.000000  3.925000e+04   \n",
       "mean    4533.774752  4.835932e+05      1.043541   -125.863580  1.473368e+12   \n",
       "std     2570.495638  9.893140e+04      0.563540      9.226500  5.506910e+09   \n",
       "min      879.000000  4.732880e+05      0.000000   -155.000000  1.463546e+12   \n",
       "25%     2790.000000  4.761230e+05      1.000000   -132.500000  1.468410e+12   \n",
       "50%     3559.000000  4.763150e+05      1.000000   -127.333333  1.473444e+12   \n",
       "75%     7456.000000  4.766090e+05      1.500000   -121.000000  1.479118e+12   \n",
       "max    11951.000000  1.747448e+06      2.000000    -64.000000  1.481143e+12   \n",
       "\n",
       "             bs_lat        bs_lng  \n",
       "count  39250.000000  39250.000000  \n",
       "mean      42.554998   -100.969685  \n",
       "std        7.722787     11.522732  \n",
       "min       36.053394   -107.463146  \n",
       "25%       39.747448   -105.074287  \n",
       "50%       39.793585   -105.043685  \n",
       "75%       39.838740   -104.998429  \n",
       "max       64.300000    -68.500000  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_mess_train.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lat</th>\n",
       "      <th>lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>39.606690</td>\n",
       "      <td>-104.958490</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>39.606690</td>\n",
       "      <td>-104.958490</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>39.637741</td>\n",
       "      <td>-104.958554</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>39.730417</td>\n",
       "      <td>-104.968940</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>39.730417</td>\n",
       "      <td>-104.968940</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         lat         lng\n",
       "0  39.606690 -104.958490\n",
       "1  39.606690 -104.958490\n",
       "2  39.637741 -104.958554\n",
       "3  39.730417 -104.968940\n",
       "4  39.730417 -104.968940"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pos_train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lat</th>\n",
       "      <th>lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>39250.000000</td>\n",
       "      <td>39250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>39.783333</td>\n",
       "      <td>-105.056343</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.206262</td>\n",
       "      <td>0.162954</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>35.770683</td>\n",
       "      <td>-107.630679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>39.708599</td>\n",
       "      <td>-105.060463</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>39.776882</td>\n",
       "      <td>-105.039440</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>39.805279</td>\n",
       "      <td>-105.018730</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>43.140387</td>\n",
       "      <td>-102.094289</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                lat           lng\n",
       "count  39250.000000  39250.000000\n",
       "mean      39.783333   -105.056343\n",
       "std        0.206262      0.162954\n",
       "min       35.770683   -107.630679\n",
       "25%       39.708599   -105.060463\n",
       "50%       39.776882   -105.039440\n",
       "75%       39.805279   -105.018730\n",
       "max       43.140387   -102.094289"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pos_train.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "listOfBs = np.union1d(np.unique(df_mess_train['bsid']), np.unique(df_mess_test['bsid'])) # determine all Base stations that received at least 1 message"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "259"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(listOfBs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Feature Matrix construction \n",
    "\n",
    "def feat_mat_const(df_mess_train, listOfBs):\n",
    "    \n",
    "    \n",
    "    return df_feat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# ground truth construction\n",
    "\n",
    "def ground_truth_const(df_mess_train, pos_train):\n",
    "\n",
    "\n",
    "    \n",
    "    return ground_truth_lat, ground_truth_lng"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df_feat = feat_mat_const(df_mess_train, listOfBs)\n",
    "df_feat.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ground_truth_lat, ground_truth_lng = ground_truth_const(df_mess_train, pos_train)\n",
    "ground_truth_lat.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def regressor_and_predict(df_feat, ground_truth_lat, ground_truth_lng, df_test):\n",
    "    \n",
    "    # train regressor and make prediction in the train set\n",
    "    # Input: df_feat: feature matrix used to train regressor\n",
    "    #        ground_truth_lat: df_feat associated latitude\n",
    "    #        ground_truth_lng: df_feat associated longitude\n",
    "    #        df_test: data frame used for prediction\n",
    "    # Output: y_pred_lat, y_pred_lng\n",
    "\n",
    "    \n",
    "    return y_pred_lat, y_pred_lng\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Make regressor and prediction using the train set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y_pred_lat, y_pred_lng = regressor_and_predict(df_feat, ground_truth_lat, ground_truth_lng, df_feat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Cross validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Evaluate result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def vincenty_vec(vec_coord):\n",
    "    vin_vec_dist = np.zeros(vec_coord.shape[0])\n",
    "    if vec_coord.shape[1] !=  4:\n",
    "        print('ERROR: Bad number of columns (shall be = 4)')\n",
    "    else:\n",
    "        vin_vec_dist = [vincenty(vec_coord[m,0:2],vec_coord[m,2:]).meters for m in range(vec_coord.shape[0])]\n",
    "    return vin_vec_dist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# evaluate distance error for each predicted point\n",
    "def Eval_geoloc(y_train_lat , y_train_lng, y_pred_lat, y_pred_lng):\n",
    "    vec_coord = np.array([y_train_lat , y_train_lng, y_pred_lat, y_pred_lng])\n",
    "    err_vec = vincenty_vec(np.transpose(vec_coord))\n",
    "    \n",
    "    return err_vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "err_vec = Eval_geoloc(ground_truth_lat , ground_truth_lng, y_pred_lat, y_pred_lng)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Plot error distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 149,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x121e7d30>"
      ]
     },
     "execution_count": 149,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAEZCAYAAABiu9n+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115a70f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "values, base = np.histogram(err_vec, bins=50000)\n",
    "cumulative = np.cumsum(values) \n",
    "plt.figure();\n",
    "plt.plot(base[:-1]/1000, cumulative / np.float(np.sum(values))  * 100.0, c='blue')\n",
    "plt.grid(); plt.xlabel('Distance Error (km)'); plt.ylabel('Cum proba (%)'); plt.axis([0, 30, 0, 100]); \n",
    "plt.title('Error Cumulative Probability'); plt.legend( [\"Opt LLR\", \"LLR 95\", \"LLR 99\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "7522.0667120795024"
      ]
     },
     "execution_count": 150,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Error criterion\n",
    "\n",
    "np.percentile(err_vec, 80)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Construct test prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 152,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>objid</th>\n",
       "      <th>bsid</th>\n",
       "      <th>did</th>\n",
       "      <th>nseq</th>\n",
       "      <th>rssi</th>\n",
       "      <th>time_ux</th>\n",
       "      <th>bs_lat</th>\n",
       "      <th>bs_lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>573be2503e952e191262c351</td>\n",
       "      <td>3578</td>\n",
       "      <td>116539</td>\n",
       "      <td>2</td>\n",
       "      <td>-111</td>\n",
       "      <td>1.463542e+12</td>\n",
       "      <td>39.728651</td>\n",
       "      <td>-105.163032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>573c05f83e952e1912758013</td>\n",
       "      <td>2617</td>\n",
       "      <td>472504</td>\n",
       "      <td>0</td>\n",
       "      <td>-136</td>\n",
       "      <td>1.463551e+12</td>\n",
       "      <td>39.779908</td>\n",
       "      <td>-105.062479</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>573c05f83e952e1912758013</td>\n",
       "      <td>3556</td>\n",
       "      <td>472504</td>\n",
       "      <td>0</td>\n",
       "      <td>-127</td>\n",
       "      <td>1.463551e+12</td>\n",
       "      <td>39.780658</td>\n",
       "      <td>-105.053676</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>573c05f83e952e1912758013</td>\n",
       "      <td>3578</td>\n",
       "      <td>472504</td>\n",
       "      <td>0</td>\n",
       "      <td>-129</td>\n",
       "      <td>1.463551e+12</td>\n",
       "      <td>39.728651</td>\n",
       "      <td>-105.163032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>573c05f83e952e1912758013</td>\n",
       "      <td>4058</td>\n",
       "      <td>472504</td>\n",
       "      <td>0</td>\n",
       "      <td>-105</td>\n",
       "      <td>1.463551e+12</td>\n",
       "      <td>39.783211</td>\n",
       "      <td>-105.088747</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                      objid  bsid     did  nseq  rssi       time_ux  \\\n",
       "0  573be2503e952e191262c351  3578  116539     2  -111  1.463542e+12   \n",
       "1  573c05f83e952e1912758013  2617  472504     0  -136  1.463551e+12   \n",
       "2  573c05f83e952e1912758013  3556  472504     0  -127  1.463551e+12   \n",
       "3  573c05f83e952e1912758013  3578  472504     0  -129  1.463551e+12   \n",
       "4  573c05f83e952e1912758013  4058  472504     0  -105  1.463551e+12   \n",
       "\n",
       "      bs_lat      bs_lng  \n",
       "0  39.728651 -105.163032  \n",
       "1  39.779908 -105.062479  \n",
       "2  39.780658 -105.053676  \n",
       "3  39.728651 -105.163032  \n",
       "4  39.783211 -105.088747  "
      ]
     },
     "execution_count": 152,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_mess_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df_feat_test = feat_mat_const(df_mess_test, listOfBs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "y_pred_lat, y_pred_lng = regressor_and_predict(df_feat, ground_truth_lat, ground_truth_lng, df_feat_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_res = pd.DataFrame(np.array([y_pred_lat, y_pred_lng]).T, columns = ['lat', 'lng'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "test_res.to_csv('pred_pos_test_list.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lat</th>\n",
       "      <th>lng</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>39.772921</td>\n",
       "      <td>-105.049617</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>39.774793</td>\n",
       "      <td>-105.062468</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>39.678762</td>\n",
       "      <td>-105.012780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>39.773710</td>\n",
       "      <td>-105.037452</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>39.678762</td>\n",
       "      <td>-105.012780</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         lat         lng\n",
       "0  39.772921 -105.049617\n",
       "1  39.774793 -105.062468\n",
       "2  39.678762 -105.012780\n",
       "3  39.773710 -105.037452\n",
       "4  39.678762 -105.012780"
      ]
     },
     "execution_count": 157,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_res.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [default]",
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}