5b. Destinations - Random Forest.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import numpy as np\n",
    "import scipy as sp\n",
    "import matplotlib as mpl\n",
    "import matplotlib.cm as cm\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "pd.set_option('display.width', 500)\n",
    "pd.set_option('display.max_columns', 100)\n",
    "pd.set_option('display.notebook_repr_html', True)\n",
    "import seaborn as sns\n",
    "sns.set_style(\"whitegrid\")\n",
    "sns.set_context(\"poster\")\n",
    "from sklearn.grid_search import GridSearchCV\n",
    "from sklearn.cross_validation import train_test_split\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.cross_validation import cross_val_score\n",
    "import geohash\n",
    "from sklearn.metrics import mean_squared_error\n",
    "import random\n",
    "import math\n",
    "import datetime\n",
    "import geopy as gp\n",
    "from geopy.distance import vincenty"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Goal: Explore the possiblity of predicting a destination (lat/long) based on the pickup location, time & day"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Note: Go directly to step 3 if you had performed 1 & 2 earlier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Load the aggregated data from the CSV file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(15285988, 5)\n"
     ]
    }
   ],
   "source": [
    "# Each line is of the format:\n",
    "\n",
    "#pickupGeohash, dropOffGeohash,time_num,day_of_week, count\n",
    "# File is also available here: https://s3.amazonaws.com/testsetu/nyc/final/groupbydestn/singlefile/part-00000\n",
    "names = [\"pickup_geohash\",\"dropoff_geohash\",\"time_num\",\"day_of_week\", \"count\"]\n",
    "df=pd.read_csv(\"./tmplocaldata/final/groupbydestn/part-00000\", header=None, names = names)\n",
    "#df = df.sample(n=10000000,weights=df['count'], replace=True)\n",
    "print df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Feature Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Get the longitude and latitude from the geohash\n",
    "def decodegeo(geo, which):\n",
    "    if len(geo) >= 6:\n",
    "        geodecoded = geohash.decode(geo)\n",
    "        return geodecoded[which]\n",
    "    else:\n",
    "        return 0\n",
    "    \n",
    "def further_data_prep(df):\n",
    "  \n",
    "    df['time_sin'] = (df['time_num'] * 2 * math.pi).apply(math.sin)\n",
    "    df['time_cos'] = (df['time_num'] * 2 * math.pi).apply(math.cos)\n",
    "    df['pickup_lat'] = df['pickup_geohash'].apply(lambda geo: decodegeo(geo, 0))\n",
    "    df['pickup_long'] = df['pickup_geohash'].apply(lambda geo: decodegeo(geo, 1))\n",
    "    df['dropoff_lat'] = df['dropoff_geohash'].apply(lambda geo: decodegeo(geo, 0))\n",
    "    df['dropoff_long'] = df['dropoff_geohash'].apply(lambda geo: decodegeo(geo, 1))\n",
    "    \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df = further_data_prep(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df.to_csv(\"./tmplocaldata/final/groupbydestn/pickup_dropoff_aggregated_with_feat_ext.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. Train-test split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 28.3 s\n"
     ]
    }
   ],
   "source": [
    "%time df=pd.read_csv(\"./tmplocaldata/final/groupbydestn/pickup_dropoff_aggregated_with_feat_ext.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Rescale longitude in the scale of latitude: so the RMSE error function in the ML algorithm works properly"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df['pickup_long_rescaled'] =  np.interp(df['pickup_long'], [-180,180], [-90,90])\n",
    "df['dropoff_long_rescaled'] =  np.interp(df['dropoff_long'], [-180,180], [-90,90])"
   ]
  },
  {
   "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>pickup_geohash</th>\n",
       "      <th>dropoff_geohash</th>\n",
       "      <th>time_num</th>\n",
       "      <th>day_of_week</th>\n",
       "      <th>count</th>\n",
       "      <th>time_sin</th>\n",
       "      <th>time_cos</th>\n",
       "      <th>pickup_lat</th>\n",
       "      <th>pickup_long</th>\n",
       "      <th>dropoff_lat</th>\n",
       "      <th>dropoff_long</th>\n",
       "      <th>pickup_long_rescaled</th>\n",
       "      <th>dropoff_long_rescaled</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>dr5rtk</td>\n",
       "      <td>dr5rsj</td>\n",
       "      <td>0.416667</td>\n",
       "      <td>4</td>\n",
       "      <td>2</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>-0.866025</td>\n",
       "      <td>40.718079</td>\n",
       "      <td>-73.943481</td>\n",
       "      <td>40.723572</td>\n",
       "      <td>-73.998413</td>\n",
       "      <td>-36.971741</td>\n",
       "      <td>-36.999207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>dr5x0z</td>\n",
       "      <td>dr5xcf</td>\n",
       "      <td>0.683333</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>-0.913545</td>\n",
       "      <td>-0.406737</td>\n",
       "      <td>40.646667</td>\n",
       "      <td>-73.789673</td>\n",
       "      <td>40.751038</td>\n",
       "      <td>-73.745728</td>\n",
       "      <td>-36.894836</td>\n",
       "      <td>-36.872864</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>dr72rd</td>\n",
       "      <td>dr782h</td>\n",
       "      <td>0.650000</td>\n",
       "      <td>3</td>\n",
       "      <td>2</td>\n",
       "      <td>-0.809017</td>\n",
       "      <td>-0.587785</td>\n",
       "      <td>40.838928</td>\n",
       "      <td>-73.844604</td>\n",
       "      <td>40.849915</td>\n",
       "      <td>-73.822632</td>\n",
       "      <td>-36.922302</td>\n",
       "      <td>-36.911316</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>dr5rvn</td>\n",
       "      <td>dr5rt5</td>\n",
       "      <td>0.083333</td>\n",
       "      <td>2</td>\n",
       "      <td>11</td>\n",
       "      <td>0.500000</td>\n",
       "      <td>0.866025</td>\n",
       "      <td>40.773010</td>\n",
       "      <td>-73.954468</td>\n",
       "      <td>40.712585</td>\n",
       "      <td>-73.954468</td>\n",
       "      <td>-36.977234</td>\n",
       "      <td>-36.977234</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>dr72jd</td>\n",
       "      <td>dr72m3</td>\n",
       "      <td>0.050000</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0.309017</td>\n",
       "      <td>0.951057</td>\n",
       "      <td>40.794983</td>\n",
       "      <td>-73.932495</td>\n",
       "      <td>40.833435</td>\n",
       "      <td>-73.943481</td>\n",
       "      <td>-36.966248</td>\n",
       "      <td>-36.971741</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  pickup_geohash dropoff_geohash  time_num  day_of_week  count  time_sin  time_cos  pickup_lat  pickup_long  dropoff_lat  dropoff_long  pickup_long_rescaled  dropoff_long_rescaled\n",
       "0         dr5rtk          dr5rsj  0.416667            4      2  0.500000 -0.866025   40.718079   -73.943481    40.723572    -73.998413            -36.971741             -36.999207\n",
       "1         dr5x0z          dr5xcf  0.683333            0      2 -0.913545 -0.406737   40.646667   -73.789673    40.751038    -73.745728            -36.894836             -36.872864\n",
       "2         dr72rd          dr782h  0.650000            3      2 -0.809017 -0.587785   40.838928   -73.844604    40.849915    -73.822632            -36.922302             -36.911316\n",
       "3         dr5rvn          dr5rt5  0.083333            2     11  0.500000  0.866025   40.773010   -73.954468    40.712585    -73.954468            -36.977234             -36.977234\n",
       "4         dr72jd          dr72m3  0.050000            0      3  0.309017  0.951057   40.794983   -73.932495    40.833435    -73.943481            -36.966248             -36.971741"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Perform the train-test split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# best practice:  random_state=42\n",
    "#trainSetSampleSize = 10000000\n",
    "testSetSize = 2000000\n",
    "testSetPosns = np.random.choice(df.shape[0],testSetSize, replace = False)\n",
    "testSet = df.iloc[testSetPosns]\n",
    "trainSet = df.drop(testSetPosns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Weighting (optional): not required as it is not able to produce a better model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#np.random.choice(a, size=2, replace = False, p=p)\n",
    "totalCount =  trainSet['count'].sum()\n",
    "trainSet[\"weight\"] = trainSet['count']/totalCount"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "trainSet = trainSet.sample(n=trainSetSampleSize,weights=trainSet['weight'], replace=True)\n",
    "#trainSet = trainSet.sample(n=trainSetSampleSize,weights=trainSet['count'], replace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Initialize the train & test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_train = trainSet[['time_num', 'time_sin', 'time_cos','day_of_week', 'pickup_lat', 'pickup_long_rescaled']]\n",
    "y_train = trainSet[['dropoff_lat', 'dropoff_long_rescaled']]\n",
    "\n",
    "X_test = testSet[['time_num', 'time_sin', 'time_cos','day_of_week', 'pickup_lat', 'pickup_long_rescaled']]\n",
    "y_test = testSet[['dropoff_lat', 'dropoff_long_rescaled']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. Machine learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Funtion for cross-validation over a grid of parameters\n",
    "\n",
    "def cv_optimize(clf, parameters, X, y, n_jobs=1, n_folds=5, score_func=None, verbose=0):\n",
    "    if score_func:\n",
    "        gs = GridSearchCV(clf, param_grid=parameters, cv=n_folds, n_jobs=n_jobs, scoring=score_func, verbose=verbose)\n",
    "    else:\n",
    "        gs = GridSearchCV(clf, param_grid=parameters, n_jobs=n_jobs, cv=n_folds, verbose=verbose)\n",
    "    gs.fit(X, y)\n",
    "    print \"BEST\", gs.best_params_, gs.best_score_, gs.grid_scores_, gs.scorer_\n",
    "    print \"Best score: \", gs.best_score_\n",
    "    best = gs.best_estimator_\n",
    "    return best"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Create a Random Forest Regression estimator\n",
    "estimator = RandomForestRegressor(n_estimators=20, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 2 folds for each of 1 candidates, totalling 2 fits\n",
      "[CV] max_features=auto, n_estimators=20, max_depth=20 ................\n",
      "[CV]  max_features=auto, n_estimators=20, max_depth=20, score=0.162738 - 3.2min\n",
      "[CV] max_features=auto, n_estimators=20, max_depth=20 ................\n",
      "[CV]  max_features=auto, n_estimators=20, max_depth=20, score=0.155466 - 3.4min"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:  3.2min\n",
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:  6.6min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "BEST {'max_features': 'auto', 'n_estimators': 20, 'max_depth': 20} 0.159101967745 [mean: 0.15910, std: 0.00364, params: {'max_features': 'auto', 'n_estimators': 20, 'max_depth': 20}] <function _passthrough_scorer at 0x000000001F0627B8>\n",
      "Best score:  0.159101967745\n",
      "Wall time: 13min 41s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# Define a grid of parameters over which to optimize the random forest\n",
    "# We will figure out which number of trees is optimal\n",
    "parameters = {\"n_estimators\": [20],\n",
    "              \"max_features\": [\"auto\"], # [\"auto\",\"sqrt\",\"log2\"]\n",
    "              \"max_depth\": [20]}\n",
    "best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=5, verbose=1)\n",
    "#best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=2, score_func='mean_squared_error', verbose=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE=  0.0817812139499\n"
     ]
    }
   ],
   "source": [
    "#No sampling + rescaled longitude: 20 estimators, 20 depth\n",
    "print \"RMSE= \",np.sqrt(mean_squared_error(best.predict(X_test),y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Optional step: re-running the above model with more trees, CV folds & depth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed: 13.5min\n",
      "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed: 67.2min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 1 candidates, totalling 5 fits\n",
      "BEST {'max_features': 'auto', 'n_estimators': 50, 'max_depth': 50} -0.00686534862405 [mean: -0.00687, std: 0.00017, params: {'max_features': 'auto', 'n_estimators': 50, 'max_depth': 50}] make_scorer(mean_squared_error, greater_is_better=False)\n",
      "Best score:  -0.00686534862405\n",
      "RMSE=  0.0812417975221\n",
      "Wall time: 1h 24min 52s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# Define a grid of parameters over which to optimize the random forest\n",
    "# We will figure out which number of trees is optimal\n",
    "parameters = {\"n_estimators\": [50],\n",
    "              \"max_features\": [\"auto\"], # [\"auto\",\"sqrt\",\"log2\"]\n",
    "              \"max_depth\": [50]}\n",
    "best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=5,score_func='mean_squared_error', verbose=1)\n",
    "#best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=2, score_func='mean_squared_error', verbose=3)\n",
    "print \"RMSE= \",np.sqrt(mean_squared_error(best.predict(X_test),y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.externals import joblib\n",
    "joblib.dump(clf, './models/dest_model.pkl') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "############# based on standard predict ################\n",
      "R^2 on training data: 0.2708\n",
      "R^2 on test data:     0.1626\n",
      "Wall time: 18min 54s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# Fit the best Random Forest and calculate R^2 values for training and test sets\n",
    "\n",
    "reg=best.fit(X_train, y_train)\n",
    "training_accuracy = reg.score(X_train, y_train)\n",
    "test_accuracy = reg.score(X_test, y_test)\n",
    "print \"############# based on standard predict ################\"\n",
    "print \"R^2 on training data: %0.4f\" % (training_accuracy)\n",
    "print \"R^2 on test data:     %0.4f\" % (test_accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE = 0.081 \n",
      "Latitude RMSE =  5.61220921856  miles\n",
      "Longitude RMSE =  8.62165474155  miles\n"
     ]
    }
   ],
   "source": [
    "# Calculate the Root Mean Squared Error\n",
    "rmse = np.sqrt(mean_squared_error(reg.predict(X_test),y_test))\n",
    "print \"RMSE = %0.3f \" % rmse\n",
    "print \"Latitude RMSE = \",rmse*69,\" miles\"\n",
    "scaledbackLongRMSE = np.interp(rmse, [-90,90], [-180,180])\n",
    "print \"Longitude RMSE = \",scaledbackLongRMSE*53,\" miles\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('pickup_lat', 0.54077211318033069),\n",
       " ('pickup_long_rescaled', 0.29392547858766693),\n",
       " ('day_of_week', 0.051206342269280024),\n",
       " ('time_sin', 0.039954843986436381),\n",
       " ('time_num', 0.039568463398994218),\n",
       " ('time_cos', 0.034572758577291481)]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# What are the most important features?\n",
    "import operator\n",
    "dict_feat_imp = dict(zip(list(X_test.columns.values),reg.feature_importances_))\n",
    "\n",
    "sorted_features = sorted(dict_feat_imp.items(), key=operator.itemgetter(1), reverse=True)\n",
    "sorted_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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>pred_dropoff_lat</th>\n",
       "      <th>pred_dropoff_long_rescaled</th>\n",
       "      <th>act_dropoff_lat</th>\n",
       "      <th>act_dropoff_long_rescaled</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>40.836495</td>\n",
       "      <td>-36.947423</td>\n",
       "      <td>40.855408</td>\n",
       "      <td>-36.949768</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>40.685634</td>\n",
       "      <td>-36.980992</td>\n",
       "      <td>40.619202</td>\n",
       "      <td>-37.004700</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>40.789750</td>\n",
       "      <td>-36.975287</td>\n",
       "      <td>40.663147</td>\n",
       "      <td>-36.900330</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>40.778937</td>\n",
       "      <td>-36.972001</td>\n",
       "      <td>40.674133</td>\n",
       "      <td>-36.999207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>40.744723</td>\n",
       "      <td>-36.964568</td>\n",
       "      <td>40.701599</td>\n",
       "      <td>-36.911316</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   pred_dropoff_lat  pred_dropoff_long_rescaled  act_dropoff_lat  act_dropoff_long_rescaled\n",
       "0         40.836495                  -36.947423        40.855408                 -36.949768\n",
       "1         40.685634                  -36.980992        40.619202                 -37.004700\n",
       "2         40.789750                  -36.975287        40.663147                 -36.900330\n",
       "3         40.778937                  -36.972001        40.674133                 -36.999207\n",
       "4         40.744723                  -36.964568        40.701599                 -36.911316"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Show some of the predictions vs actual\n",
    "#actualPredictedDf = pd.DataFrame(np.hstack(y_test, reg.predict(X_test), names=[\"act_dropoff_lat\", \"act_dropoff_lat_rescaled\",\"pred_dropoff_lat\", \"pred_dropoff_lat_rescaled\"])\n",
    "\n",
    "actualPredictedDf = pd.DataFrame(np.hstack((reg.predict(X_test), y_test.values)),columns=[\"pred_dropoff_lat\", \"pred_dropoff_long_rescaled\", \"act_dropoff_lat\", \"act_dropoff_long_rescaled\"])\n",
    "actualPredictedDf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Scaleback the longitudes to original scale"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>pred_dropoff_lat</th>\n",
       "      <th>act_dropoff_lat</th>\n",
       "      <th>pred_dropoff_long</th>\n",
       "      <th>act_dropoff_long</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>40.836495</td>\n",
       "      <td>40.855408</td>\n",
       "      <td>-73.894847</td>\n",
       "      <td>-73.899536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>40.685634</td>\n",
       "      <td>40.619202</td>\n",
       "      <td>-73.961985</td>\n",
       "      <td>-74.009399</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>40.789750</td>\n",
       "      <td>40.663147</td>\n",
       "      <td>-73.950575</td>\n",
       "      <td>-73.800659</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>40.778937</td>\n",
       "      <td>40.674133</td>\n",
       "      <td>-73.944002</td>\n",
       "      <td>-73.998413</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>40.744723</td>\n",
       "      <td>40.701599</td>\n",
       "      <td>-73.929136</td>\n",
       "      <td>-73.822632</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   pred_dropoff_lat  act_dropoff_lat  pred_dropoff_long  act_dropoff_long\n",
       "0         40.836495        40.855408         -73.894847        -73.899536\n",
       "1         40.685634        40.619202         -73.961985        -74.009399\n",
       "2         40.789750        40.663147         -73.950575        -73.800659\n",
       "3         40.778937        40.674133         -73.944002        -73.998413\n",
       "4         40.744723        40.701599         -73.929136        -73.822632"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "actualPredictedDf['pred_dropoff_long'] =  np.interp(actualPredictedDf['pred_dropoff_long_rescaled'],[-90,90], [-180,180])\n",
    "actualPredictedDf['act_dropoff_long'] =  np.interp(actualPredictedDf['act_dropoff_long_rescaled'], [-90,90], [-180,180])\n",
    "actualPredictedDf.drop(\"pred_dropoff_long_rescaled\", inplace=True, axis=1)\n",
    "actualPredictedDf.drop(\"act_dropoff_long_rescaled\", inplace=True , axis=1)\n",
    "actualPredictedDf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Evaluate performance of model based on scaled back longitude data (to original range)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def getLatLongBasedOnGeocode(lat, longitude, accuracy = 6):\n",
    "    gh = geohash.encode(lat, longitude, accuracy)\n",
    "    ghLatLong = geohash.decode(gh)\n",
    "    return ghLatLong[0], ghLatLong[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(40.74554443359375, -73.9215087890625)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "getLatLongBasedOnGeocode(40.745861, -73.917152)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "actualPredictedDf[\"pred_dropoff_lat_from_gh\"], actualPredictedDf[\"pred_dropoff_long_from_gh\"] = zip(*actualPredictedDf.apply(lambda x: getLatLongBasedOnGeocode(x['pred_dropoff_lat'], x['pred_dropoff_long'], 6), axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "actualPredictedDf[\"act_dropoff_lat_from_gh\"], actualPredictedDf[\"act_dropoff_long_from_gh\"] = zip(*actualPredictedDf.apply(lambda x: getLatLongBasedOnGeocode(x['act_dropoff_lat'], x['act_dropoff_long'], 6), axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "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>act_dropoff_lat</th>\n",
       "      <th>act_dropoff_long</th>\n",
       "      <th>act_dropoff_lat_from_gh</th>\n",
       "      <th>act_dropoff_long_from_gh</th>\n",
       "      <th>pred_dropoff_lat</th>\n",
       "      <th>pred_dropoff_long</th>\n",
       "      <th>pred_dropoff_lat_from_gh</th>\n",
       "      <th>pred_dropoff_long_from_gh</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>40.855408</td>\n",
       "      <td>-73.899536</td>\n",
       "      <td>40.855408</td>\n",
       "      <td>-73.899536</td>\n",
       "      <td>40.836495</td>\n",
       "      <td>-73.894847</td>\n",
       "      <td>40.838928</td>\n",
       "      <td>-73.899536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>40.619202</td>\n",
       "      <td>-74.009399</td>\n",
       "      <td>40.619202</td>\n",
       "      <td>-74.009399</td>\n",
       "      <td>40.685634</td>\n",
       "      <td>-73.961985</td>\n",
       "      <td>40.685120</td>\n",
       "      <td>-73.965454</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>40.663147</td>\n",
       "      <td>-73.800659</td>\n",
       "      <td>40.663147</td>\n",
       "      <td>-73.800659</td>\n",
       "      <td>40.789750</td>\n",
       "      <td>-73.950575</td>\n",
       "      <td>40.789490</td>\n",
       "      <td>-73.954468</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>40.674133</td>\n",
       "      <td>-73.998413</td>\n",
       "      <td>40.674133</td>\n",
       "      <td>-73.998413</td>\n",
       "      <td>40.778937</td>\n",
       "      <td>-73.944002</td>\n",
       "      <td>40.778503</td>\n",
       "      <td>-73.943481</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>40.701599</td>\n",
       "      <td>-73.822632</td>\n",
       "      <td>40.701599</td>\n",
       "      <td>-73.822632</td>\n",
       "      <td>40.744723</td>\n",
       "      <td>-73.929136</td>\n",
       "      <td>40.745544</td>\n",
       "      <td>-73.932495</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   act_dropoff_lat  act_dropoff_long  act_dropoff_lat_from_gh  act_dropoff_long_from_gh  pred_dropoff_lat  pred_dropoff_long  pred_dropoff_lat_from_gh  pred_dropoff_long_from_gh\n",
       "0        40.855408        -73.899536                40.855408                -73.899536         40.836495         -73.894847                 40.838928                 -73.899536\n",
       "1        40.619202        -74.009399                40.619202                -74.009399         40.685634         -73.961985                 40.685120                 -73.965454\n",
       "2        40.663147        -73.800659                40.663147                -73.800659         40.789750         -73.950575                 40.789490                 -73.954468\n",
       "3        40.674133        -73.998413                40.674133                -73.998413         40.778937         -73.944002                 40.778503                 -73.943481\n",
       "4        40.701599        -73.822632                40.701599                -73.822632         40.744723         -73.929136                 40.745544                 -73.932495"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "actualPredictedDf = actualPredictedDf[[\"act_dropoff_lat\",\"act_dropoff_long\",\"act_dropoff_lat_from_gh\",\"act_dropoff_long_from_gh\",\"pred_dropoff_lat\",\"pred_dropoff_long\",\"pred_dropoff_lat_from_gh\",\"pred_dropoff_long_from_gh\" ]]\n",
    "actualPredictedDf.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Compute RMSE based on raw lat/long"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Raw Lat RMSE:  0.0973409971659\n",
      "Raw Long RMSE:  0.122571431775\n",
      "Lat RMSE in miles =  6.71652880444  miles\n",
      "Long RMSE in miles =  6.49628588405  miles\n"
     ]
    }
   ],
   "source": [
    "#rmse = np.sqrt(mean_squared_error(reg.predict(Xtest),ytest))\n",
    "latRMSE = np.sqrt(mean_squared_error(actualPredictedDf[\"act_dropoff_lat\"] , actualPredictedDf[\"pred_dropoff_lat\"]))\n",
    "longRMSE = np.sqrt(mean_squared_error(actualPredictedDf[\"act_dropoff_long\"] , actualPredictedDf[\"pred_dropoff_long\"]))\n",
    "print \"Raw Lat RMSE: \", latRMSE\n",
    "print \"Raw Long RMSE: \", longRMSE\n",
    "print \"Lat RMSE in miles = \",latRMSE*69,\" miles\"\n",
    "print \"Long RMSE in miles = \",longRMSE*53,\" miles\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Compute RMSE based geohashed lat/long"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lat RMSE(geohashed):  0.0973553068517\n",
      "Long RMSE(geohashed):  0.122609611492\n",
      "Lat RMSE in miles(geohashed) =  6.71751617277  miles\n",
      "Long RMSE in miles(geohashed) =  6.49830940906  miles\n"
     ]
    }
   ],
   "source": [
    "#rmse = np.sqrt(mean_squared_error(reg.predict(Xtest),ytest))\n",
    "latRMSE_GH = np.sqrt(mean_squared_error(actualPredictedDf[\"act_dropoff_lat_from_gh\"] , actualPredictedDf[\"pred_dropoff_lat_from_gh\"]))\n",
    "longRMSE_GH = np.sqrt(mean_squared_error(actualPredictedDf[\"act_dropoff_long_from_gh\"] , actualPredictedDf[\"pred_dropoff_long_from_gh\"]))\n",
    "print \"Lat RMSE(geohashed): \", latRMSE_GH\n",
    "print \"Long RMSE(geohashed): \", longRMSE_GH\n",
    "print \"Lat RMSE in miles(geohashed) = \",latRMSE_GH*69,\" miles\"\n",
    "print \"Long RMSE in miles(geohashed) = \",longRMSE_GH*53,\" miles\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Compute physical distance between actual and predicted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "#compute actual error distance\n",
    "actualPredictedDf[\"error_in_miles\"] = actualPredictedDf.apply(lambda row: vincenty((row[\"act_dropoff_lat_from_gh\"], row[\"act_dropoff_long_from_gh\"]), (row[\"pred_dropoff_lat_from_gh\"], row[\"pred_dropoff_long_from_gh\"])).miles, axis=1)\n",
    "#actualPredictedDf[[\"act_dropoff_lat\", \"act_dropoff_long\", \"pred_dropoff_lat_from_gh\", \"pred_dropoff_long_from_gh\"]].apply(lambda x: vincenty((x[\"act_dropoff_lat\"], x[1]), (x[2], x[3])).miles, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Avg distance error is:  4.77248220597  miles\n"
     ]
    }
   ],
   "source": [
    "avgDistanceError = actualPredictedDf[\"error_in_miles\"].mean()\n",
    "print \"Avg distance error is: \", avgDistanceError, \" miles\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Analysis of errors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7aa53588>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAygAAAIbCAYAAAAAbbpGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKCgAAAAAbOP/AH4JWy0HkV0VAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7aa530f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "actualPredictedDf[\"error_in_miles\"].hist(bins=1000) #, range =(0,200))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "actualPredictedDf.to_csv(\"./tmplocaldata/final/groupbydestn/destination_prediction.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Reference [link](http://geography.about.com/library/faq/blqzdistancedegree.htm)\n",
    "Each degree of latitude is approximately 69 miles (111 kilometers) apart. The range varies (due to the earth's slightly ellipsoid shape) from 68.703 miles (110.567 km) at the equator to 69.407 (111.699 km) at the poles. \n",
    "A degree of longitude is widest at the equator at 69.172 miles (111.321) and gradually shrinks to zero at the poles. At 40° north or south the distance between a degree of longitude is 53 miles (85 km)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Appendix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## With sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.5s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.5s finished\n",
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:    2.0s\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    2.0s finished\n",
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:    0.4s\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.4s finished\n",
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:    1.9s\n",
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    1.9s finished\n",
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:    0.4s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "building tree 1 of 1\n",
      " R^2 (train) = 0.234, R^2 (valid) = -0.057, RMSE (train) = 0.036, RMSE (valid) = 0.128"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.4s finished\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.2s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.2s finished\n",
      "[Parallel(n_jobs=2)]: Done   1 out of   2 | elapsed:    2.0s remaining:    2.0s\n",
      "[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    2.0s finished\n",
      "[Parallel(n_jobs=2)]: Done   1 out of   2 | elapsed:    0.5s remaining:    0.5s\n",
      "[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    0.5s finished\n",
      "[Parallel(n_jobs=2)]: Done   1 out of   2 | elapsed:    2.1s remaining:    2.1s\n",
      "[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    2.1s finished\n",
      "[Parallel(n_jobs=2)]: Done   1 out of   2 | elapsed:    0.5s remaining:    0.5s\n",
      "[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:    0.5s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.275, R^2 (valid) = 0.016, RMSE (train) = 0.035, RMSE (valid) = 0.123"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.4s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.4s finished\n",
      "[Parallel(n_jobs=3)]: Done   2 out of   3 | elapsed:    2.2s remaining:    1.1s\n",
      "[Parallel(n_jobs=3)]: Done   1 out of   3 | elapsed:    2.3s remaining:    4.7s\n",
      "[Parallel(n_jobs=3)]: Done   3 out of   3 | elapsed:    2.3s finished\n",
      "[Parallel(n_jobs=3)]: Done   1 out of   3 | elapsed:    0.5s remaining:    1.2s\n",
      "[Parallel(n_jobs=3)]: Done   2 out of   3 | elapsed:    0.5s remaining:    0.2s\n",
      "[Parallel(n_jobs=3)]: Done   3 out of   3 | elapsed:    0.6s finished\n",
      "[Parallel(n_jobs=3)]: Done   2 out of   3 | elapsed:    2.2s remaining:    1.1s\n",
      "[Parallel(n_jobs=3)]: Done   1 out of   3 | elapsed:    2.3s remaining:    4.7s\n",
      "[Parallel(n_jobs=3)]: Done   3 out of   3 | elapsed:    2.3s finished\n",
      "[Parallel(n_jobs=3)]: Done   2 out of   3 | elapsed:    0.5s remaining:    0.2s\n",
      "[Parallel(n_jobs=3)]: Done   1 out of   3 | elapsed:    0.6s remaining:    1.3s\n",
      "[Parallel(n_jobs=3)]: Done   3 out of   3 | elapsed:    0.6s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.283, R^2 (valid) = 0.035, RMSE (train) = 0.035, RMSE (valid) = 0.122"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   42.0s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   42.0s finished\n",
      "[Parallel(n_jobs=4)]: Done   1 out of   4 | elapsed:    2.7s remaining:    8.3s\n",
      "[Parallel(n_jobs=4)]: Done   2 out of   4 | elapsed:    2.7s remaining:    2.7s\n",
      "[Parallel(n_jobs=4)]: Done   4 out of   4 | elapsed:    2.7s finished\n",
      "[Parallel(n_jobs=4)]: Done   2 out of   4 | elapsed:    0.6s remaining:    0.6s\n",
      "[Parallel(n_jobs=4)]: Done   1 out of   4 | elapsed:    0.7s remaining:    2.3s\n",
      "[Parallel(n_jobs=4)]: Done   4 out of   4 | elapsed:    0.7s finished\n",
      "[Parallel(n_jobs=4)]: Done   2 out of   4 | elapsed:    2.5s remaining:    2.5s\n",
      "[Parallel(n_jobs=4)]: Done   1 out of   4 | elapsed:    2.6s remaining:    7.9s\n",
      "[Parallel(n_jobs=4)]: Done   4 out of   4 | elapsed:    2.6s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.286, R^2 (valid) = 0.055, RMSE (train) = 0.035, RMSE (valid) = 0.121"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=4)]: Done   2 out of   4 | elapsed:    0.7s remaining:    0.7s\n",
      "[Parallel(n_jobs=4)]: Done   1 out of   4 | elapsed:    0.7s remaining:    2.3s\n",
      "[Parallel(n_jobs=4)]: Done   4 out of   4 | elapsed:    0.7s finished\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.2s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.2s finished\n",
      "[Parallel(n_jobs=5)]: Done   1 out of   5 | elapsed:    2.9s remaining:   11.8s\n",
      "[Parallel(n_jobs=5)]: Done   3 out of   5 | elapsed:    2.9s remaining:    1.9s\n",
      "[Parallel(n_jobs=5)]: Done   5 out of   5 | elapsed:    3.0s finished\n",
      "[Parallel(n_jobs=5)]: Done   3 out of   5 | elapsed:    0.7s remaining:    0.5s\n",
      "[Parallel(n_jobs=5)]: Done   1 out of   5 | elapsed:    0.8s remaining:    3.4s\n",
      "[Parallel(n_jobs=5)]: Done   5 out of   5 | elapsed:    0.8s finished\n",
      "[Parallel(n_jobs=5)]: Done   3 out of   5 | elapsed:    2.9s remaining:    1.9s\n",
      "[Parallel(n_jobs=5)]: Done   1 out of   5 | elapsed:    3.0s remaining:   12.3s\n",
      "[Parallel(n_jobs=5)]: Done   5 out of   5 | elapsed:    3.0s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.286, R^2 (valid) = 0.068, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=5)]: Done   3 out of   5 | elapsed:    0.7s remaining:    0.5s\n",
      "[Parallel(n_jobs=5)]: Done   1 out of   5 | elapsed:    0.7s remaining:    3.3s\n",
      "[Parallel(n_jobs=5)]: Done   5 out of   5 | elapsed:    0.7s finished\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s finished\n",
      "[Parallel(n_jobs=6)]: Done   2 out of   6 | elapsed:    2.8s remaining:    5.7s\n",
      "[Parallel(n_jobs=6)]: Done   4 out of   6 | elapsed:    2.9s remaining:    1.4s\n",
      "[Parallel(n_jobs=6)]: Done   1 out of   6 | elapsed:    3.0s remaining:   15.5s\n",
      "[Parallel(n_jobs=6)]: Done   6 out of   6 | elapsed:    3.1s finished\n",
      "[Parallel(n_jobs=6)]: Done   4 out of   6 | elapsed:    0.7s remaining:    0.3s\n",
      "[Parallel(n_jobs=6)]: Done   2 out of   6 | elapsed:    0.7s remaining:    1.5s\n",
      "[Parallel(n_jobs=6)]: Done   1 out of   6 | elapsed:    0.8s remaining:    4.4s\n",
      "[Parallel(n_jobs=6)]: Done   6 out of   6 | elapsed:    0.9s finished\n",
      "[Parallel(n_jobs=6)]: Done   4 out of   6 | elapsed:    2.9s remaining:    1.4s\n",
      "[Parallel(n_jobs=6)]: Done   1 out of   6 | elapsed:    3.2s remaining:   16.5s\n",
      "[Parallel(n_jobs=6)]: Done   2 out of   6 | elapsed:    3.2s remaining:    6.6s\n",
      "[Parallel(n_jobs=6)]: Done   6 out of   6 | elapsed:    3.3s finished\n",
      "[Parallel(n_jobs=6)]: Done   1 out of   6 | elapsed:    0.7s remaining:    4.0s\n",
      "[Parallel(n_jobs=6)]: Done   4 out of   6 | elapsed:    0.7s remaining:    0.3s\n",
      "[Parallel(n_jobs=6)]: Done   2 out of   6 | elapsed:    0.8s remaining:    1.7s\n",
      "[Parallel(n_jobs=6)]: Done   6 out of   6 | elapsed:    0.8s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.288, R^2 (valid) = 0.068, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.9s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.9s finished\n",
      "[Parallel(n_jobs=7)]: Done   3 out of   7 | elapsed:    3.4s remaining:    4.6s\n",
      "[Parallel(n_jobs=7)]: Done   1 out of   7 | elapsed:    3.4s remaining:   21.2s\n",
      "[Parallel(n_jobs=7)]: Done   5 out of   7 | elapsed:    3.5s remaining:    1.3s\n",
      "[Parallel(n_jobs=7)]: Done   7 out of   7 | elapsed:    3.5s finished\n",
      "[Parallel(n_jobs=7)]: Done   3 out of   7 | elapsed:    0.8s remaining:    1.1s\n",
      "[Parallel(n_jobs=7)]: Done   5 out of   7 | elapsed:    0.9s remaining:    0.3s\n",
      "[Parallel(n_jobs=7)]: Done   1 out of   7 | elapsed:    0.9s remaining:    5.9s\n",
      "[Parallel(n_jobs=7)]: Done   7 out of   7 | elapsed:    0.9s finished\n",
      "[Parallel(n_jobs=7)]: Done   3 out of   7 | elapsed:    3.4s remaining:    4.5s\n",
      "[Parallel(n_jobs=7)]: Done   1 out of   7 | elapsed:    3.4s remaining:   20.8s\n",
      "[Parallel(n_jobs=7)]: Done   5 out of   7 | elapsed:    3.4s remaining:    1.3s\n",
      "[Parallel(n_jobs=7)]: Done   7 out of   7 | elapsed:    3.4s finished\n",
      "[Parallel(n_jobs=7)]: Done   5 out of   7 | elapsed:    0.8s remaining:    0.3s\n",
      "[Parallel(n_jobs=7)]: Done   3 out of   7 | elapsed:    0.9s remaining:    1.2s\n",
      "[Parallel(n_jobs=7)]: Done   1 out of   7 | elapsed:    0.9s remaining:    6.0s\n",
      "[Parallel(n_jobs=7)]: Done   7 out of   7 | elapsed:    0.9s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.288, R^2 (valid) = 0.069, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.5s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   7 out of   8 | elapsed:    3.5s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done   4 out of   8 | elapsed:    3.9s remaining:    3.9s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   8 | elapsed:    4.2s remaining:   30.3s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   8 | elapsed:    4.2s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of   8 | elapsed:    0.9s remaining:    0.9s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   8 | elapsed:    1.0s remaining:    7.5s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of   8 | elapsed:    1.0s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   8 | elapsed:    1.0s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of   8 | elapsed:    3.5s remaining:    3.5s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   8 | elapsed:    3.9s remaining:   27.7s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of   8 | elapsed:    3.9s remaining:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   8 | elapsed:    3.9s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of   8 | elapsed:    0.9s remaining:    0.9s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   8 | elapsed:    0.9s remaining:    7.2s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of   8 | elapsed:    1.0s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   8 | elapsed:    1.0s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.289, R^2 (valid) = 0.069, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   9 | elapsed:    3.7s remaining:   30.7s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   9 | elapsed:    3.9s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done   2 out of   9 | elapsed:    3.9s remaining:   13.8s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of   9 | elapsed:    3.9s remaining:    3.1s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of   9 | elapsed:    5.9s finished\n",
      "[Parallel(n_jobs=8)]: Done   2 out of   9 | elapsed:    1.0s remaining:    3.6s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of   9 | elapsed:    1.0s remaining:    0.8s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   9 | elapsed:    1.0s remaining:    0.0s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   9 | elapsed:    1.0s remaining:    8.6s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of   9 | elapsed:    1.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   2 out of   9 | elapsed:    3.7s remaining:   13.1s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   9 | elapsed:    3.8s remaining:   31.1s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of   9 | elapsed:    3.8s remaining:    3.0s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   9 | elapsed:    3.9s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of   9 | elapsed:    5.9s finished\n",
      "[Parallel(n_jobs=8)]: Done   2 out of   9 | elapsed:    1.0s remaining:    3.8s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of   9 | elapsed:    1.0s remaining:    0.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of   9 | elapsed:    1.2s remaining:   10.0s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of   9 | elapsed:    1.2s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of   9 | elapsed:    1.7s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.289, R^2 (valid) = 0.067, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.8s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  10 | elapsed:    3.9s remaining:    2.5s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  10 | elapsed:    4.0s remaining:   36.4s\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  10 | elapsed:    4.2s remaining:    9.9s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  10 | elapsed:    6.0s remaining:    0.6s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  10 | elapsed:    6.2s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  10 | elapsed:    0.9s remaining:    0.6s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  10 | elapsed:    0.9s remaining:    9.3s\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  10 | elapsed:    1.2s remaining:    3.0s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  10 | elapsed:    1.6s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  10 | elapsed:    1.6s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  10 | elapsed:    3.8s remaining:    9.1s\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  10 | elapsed:    4.0s remaining:    2.6s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  10 | elapsed:    4.0s remaining:   37.0s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  10 | elapsed:    6.1s remaining:    0.6s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  10 | elapsed:    6.1s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  10 | elapsed:    0.9s remaining:    2.3s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  10 | elapsed:    1.0s remaining:    9.8s\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  10 | elapsed:    1.1s remaining:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  10 | elapsed:    1.5s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  10 | elapsed:    1.6s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.289, R^2 (valid) = 0.065, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.5s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  11 | elapsed:    3.5s remaining:    6.1s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  11 | elapsed:    3.7s remaining:   38.4s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  11 | elapsed:    3.8s remaining:    2.1s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  11 | elapsed:    6.0s remaining:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  11 | elapsed:    6.1s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  11 | elapsed:    0.9s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  11 | elapsed:    0.9s remaining:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  11 | elapsed:    1.0s remaining:   10.4s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  11 | elapsed:    1.6s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  11 | elapsed:    1.6s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  11 | elapsed:    3.8s remaining:   39.1s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  11 | elapsed:    3.8s remaining:    2.1s\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  11 | elapsed:    3.8s remaining:    6.8s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  11 | elapsed:    6.2s remaining:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  11 | elapsed:    6.2s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  11 | elapsed:    0.8s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  11 | elapsed:    0.9s remaining:   10.1s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  11 | elapsed:    1.0s remaining:    0.5s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  11 | elapsed:    1.6s remaining:    0.1s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  11 | elapsed:    1.6s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.290, R^2 (valid) = 0.063, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  12 | elapsed:    3.7s remaining:    3.7s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  12 | elapsed:    3.8s remaining:   42.5s\n",
      "[Parallel(n_jobs=8)]: Done   2 out of  12 | elapsed:    3.8s remaining:   19.3s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  12 | elapsed:    6.4s remaining:    1.2s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  12 | elapsed:    6.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  12 | elapsed:    0.9s remaining:   11.2s\n",
      "[Parallel(n_jobs=8)]: Done   2 out of  12 | elapsed:    1.0s remaining:    5.2s\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  12 | elapsed:    1.0s remaining:    1.0s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  12 | elapsed:    1.7s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  12 | elapsed:    1.7s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  12 | elapsed:    3.6s remaining:    3.6s\n",
      "[Parallel(n_jobs=8)]: Done   2 out of  12 | elapsed:    3.8s remaining:   19.5s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  12 | elapsed:    3.9s remaining:   43.4s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  12 | elapsed:    6.4s remaining:    1.2s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  12 | elapsed:    6.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   2 out of  12 | elapsed:    0.8s remaining:    4.6s\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  12 | elapsed:    1.0s remaining:    1.0s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  12 | elapsed:    1.0s remaining:   12.4s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  12 | elapsed:    1.6s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  12 | elapsed:    1.7s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.290, R^2 (valid) = 0.063, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.7s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.7s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  13 | elapsed:    3.7s remaining:   12.7s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  13 | elapsed:    3.8s remaining:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  13 | elapsed:    4.0s remaining:   48.9s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  13 | elapsed:    6.8s remaining:    1.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  13 | elapsed:    6.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  13 | elapsed:    0.9s remaining:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  13 | elapsed:    1.0s remaining:    0.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  13 | elapsed:    1.0s remaining:   12.5s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  13 | elapsed:    1.7s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  13 | elapsed:    1.7s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  13 | elapsed:    3.8s remaining:   46.4s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  13 | elapsed:    3.8s remaining:    3.3s\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  13 | elapsed:    3.9s remaining:   13.1s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  13 | elapsed:    6.8s remaining:    1.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  13 | elapsed:    6.9s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  13 | elapsed:    0.9s remaining:    3.2s\n",
      "[Parallel(n_jobs=8)]: Done   7 out of  13 | elapsed:    0.9s remaining:    0.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  13 | elapsed:    1.0s remaining:   13.2s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  13 | elapsed:    1.7s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  13 | elapsed:    1.8s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.062, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.0s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.0s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  14 | elapsed:    3.7s remaining:    9.4s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  14 | elapsed:    3.8s remaining:    2.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  14 | elapsed:    3.8s remaining:   51.1s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  14 | elapsed:    6.9s remaining:    1.1s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  14 | elapsed:    7.0s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  14 | elapsed:    0.9s remaining:    2.3s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  14 | elapsed:    0.9s remaining:   13.1s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  14 | elapsed:    1.0s remaining:    0.7s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  14 | elapsed:    1.7s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  14 | elapsed:    1.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  14 | elapsed:    3.8s remaining:    9.5s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  14 | elapsed:    3.8s remaining:    2.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  14 | elapsed:    3.8s remaining:   51.2s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  14 | elapsed:    6.8s remaining:    1.1s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  14 | elapsed:    7.0s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  14 | elapsed:    0.8s remaining:    2.2s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  14 | elapsed:    1.0s remaining:   14.0s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  14 | elapsed:    1.0s remaining:    0.8s\n",
      "[Parallel(n_jobs=8)]: Done  12 out of  14 | elapsed:    1.8s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  14 | elapsed:    1.8s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.063, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   40.8s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   40.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  15 | elapsed:    3.7s remaining:   52.7s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  15 | elapsed:    3.8s remaining:    7.6s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  15 | elapsed:    7.1s remaining:    4.7s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  15 | elapsed:    7.1s remaining:    1.0s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  15 | elapsed:    7.2s finished\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  15 | elapsed:    1.0s remaining:    2.1s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  15 | elapsed:    1.0s remaining:   15.1s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  15 | elapsed:    1.8s remaining:    1.1s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  15 | elapsed:    1.9s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  15 | elapsed:    1.9s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  15 | elapsed:    3.8s remaining:   54.4s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  15 | elapsed:    3.9s remaining:    7.9s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  15 | elapsed:    7.3s remaining:    1.0s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  15 | elapsed:    7.3s remaining:    4.9s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  15 | elapsed:    7.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  15 | elapsed:    1.0s remaining:    2.2s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  15 | elapsed:    1.1s remaining:   16.1s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  15 | elapsed:    1.8s remaining:    1.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  15 | elapsed:    1.9s remaining:    0.2s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  15 | elapsed:    1.9s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.063, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.1s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   41.1s finished\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  16 | elapsed:    3.8s remaining:    3.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  16 | elapsed:    3.9s remaining:  1.0min\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  16 | elapsed:    4.1s remaining:   18.1s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  16 | elapsed:    7.5s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  16 | elapsed:    7.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  16 | elapsed:    1.0s remaining:    4.6s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  16 | elapsed:    1.0s remaining:   16.1s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  16 | elapsed:    1.0s remaining:    1.0s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  16 | elapsed:    1.9s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  16 | elapsed:    2.0s finished\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  16 | elapsed:    3.7s remaining:   16.6s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  16 | elapsed:    3.8s remaining:    3.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  16 | elapsed:    3.9s remaining:   59.2s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  16 | elapsed:    7.4s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  16 | elapsed:    7.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  16 | elapsed:    0.9s remaining:   15.1s\n",
      "[Parallel(n_jobs=8)]: Done   3 out of  16 | elapsed:    0.9s remaining:    4.4s\n",
      "[Parallel(n_jobs=8)]: Done   8 out of  16 | elapsed:    1.1s remaining:    1.1s\n",
      "[Parallel(n_jobs=8)]: Done  13 out of  16 | elapsed:    1.9s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  16 | elapsed:    1.9s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.065, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.4s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  17 | elapsed:    3.5s remaining:   11.7s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  17 | elapsed:    3.8s remaining:  1.0min\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  17 | elapsed:    7.1s remaining:    6.3s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  17 | elapsed:    7.5s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  17 out of  17 | elapsed:    9.3s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  17 | elapsed:    0.8s remaining:    2.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  17 | elapsed:    1.0s remaining:   18.2s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  17 | elapsed:    1.8s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  17 | elapsed:    1.9s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  17 out of  17 | elapsed:    2.4s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  17 | elapsed:    3.6s remaining:   11.8s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  17 | elapsed:    3.7s remaining:  1.0min\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  17 | elapsed:    7.1s remaining:    6.3s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  17 | elapsed:    7.6s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  17 out of  17 | elapsed:    9.3s finished\n",
      "[Parallel(n_jobs=8)]: Done   4 out of  17 | elapsed:    0.9s remaining:    3.1s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  17 | elapsed:    1.2s remaining:   20.1s\n",
      "[Parallel(n_jobs=8)]: Done   9 out of  17 | elapsed:    1.8s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done  14 out of  17 | elapsed:    2.1s remaining:    0.4s\n",
      "[Parallel(n_jobs=8)]: Done  17 out of  17 | elapsed:    2.4s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.065, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.1s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   39.1s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  18 | elapsed:    3.9s remaining:  1.1min\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  18 | elapsed:    4.0s remaining:   10.7s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  18 | elapsed:    7.6s remaining:    6.1s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  18 | elapsed:    7.9s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  18 out of  18 | elapsed:    9.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  18 | elapsed:    0.9s remaining:   16.8s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  18 | elapsed:    1.0s remaining:    2.7s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  18 | elapsed:    1.9s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  18 | elapsed:    2.0s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  18 out of  18 | elapsed:    2.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  18 | elapsed:    3.9s remaining:   10.2s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  18 | elapsed:    3.9s remaining:  1.1min\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  18 | elapsed:    7.5s remaining:    6.0s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  18 | elapsed:    7.9s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  18 out of  18 | elapsed:    9.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  18 | elapsed:    1.1s remaining:   19.8s\n",
      "[Parallel(n_jobs=8)]: Done   5 out of  18 | elapsed:    1.1s remaining:    2.9s\n",
      "[Parallel(n_jobs=8)]: Done  10 out of  18 | elapsed:    2.1s remaining:    1.6s\n",
      "[Parallel(n_jobs=8)]: Done  15 out of  18 | elapsed:    2.1s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  18 out of  18 | elapsed:    2.7s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.065, RMSE (train) = 0.035, RMSE (valid) = 0.120"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   40.1s remaining:    0.0s\n",
      "[Parallel(n_jobs=-1)]: Done   1 out of   1 | elapsed:   40.1s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  19 | elapsed:    3.9s remaining:    8.6s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  19 | elapsed:    3.9s remaining:  1.2min\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  19 | elapsed:    7.5s remaining:    5.5s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  19 | elapsed:    8.0s remaining:    1.4s\n",
      "[Parallel(n_jobs=8)]: Done  19 out of  19 | elapsed:   10.3s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  19 | elapsed:    1.1s remaining:    2.4s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  19 | elapsed:    1.1s remaining:   20.9s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  19 | elapsed:    2.1s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  19 | elapsed:    2.3s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  19 out of  19 | elapsed:    2.8s finished\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  19 | elapsed:    3.8s remaining:    8.4s\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  19 | elapsed:    3.9s remaining:  1.2min\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  19 | elapsed:    7.9s remaining:    5.8s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  19 | elapsed:    8.3s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  19 out of  19 | elapsed:   10.5s finished\n",
      "[Parallel(n_jobs=8)]: Done   1 out of  19 | elapsed:    0.9s remaining:   18.7s\n",
      "[Parallel(n_jobs=8)]: Done   6 out of  19 | elapsed:    1.1s remaining:    2.6s\n",
      "[Parallel(n_jobs=8)]: Done  11 out of  19 | elapsed:    2.2s remaining:    1.5s\n",
      "[Parallel(n_jobs=8)]: Done  16 out of  19 | elapsed:    2.3s remaining:    0.3s\n",
      "[Parallel(n_jobs=8)]: Done  19 out of  19 | elapsed:    2.8s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "building tree 1 of 1\n",
      " R^2 (train) = 0.291, R^2 (valid) = 0.062, RMSE (train) = 0.035, RMSE (valid) = 0.120\n"
     ]
    }
   ],
   "source": [
    "reg = RandomForestRegressor(n_estimators=1, max_depth=20, n_jobs=-1, verbose=4, warm_start=True)\n",
    "for n in range(1,20):\n",
    "   reg.set_params(n_estimators=n)\n",
    "   reg.fit(X_train,y_train) #, sample_weight=pickup_count_train.values)\n",
    "   training_accuracy = reg.score(X_train, y_train)\n",
    "   valid_accuracy = reg.score(X_test, y_test)\n",
    "   rmsetrain = np.sqrt(mean_squared_error(reg.predict(X_train),y_train))\n",
    "   rmsevalid = np.sqrt(mean_squared_error(reg.predict(X_test),y_test))\n",
    "   print \" R^2 (train) = %0.3f, R^2 (valid) = %0.3f, RMSE (train) = %0.3f, RMSE (valid) = %0.3f\" % (training_accuracy, valid_accuracy, rmsetrain, rmsevalid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE=  0.118465757808\n"
     ]
    }
   ],
   "source": [
    "#With weighted sampling: 20 estimators, 20 depth\n",
    "print \"RMSE= \",np.sqrt(mean_squared_error(best.predict(X_test),y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Without sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 2 folds for each of 1 candidates, totalling 2 fits\n",
      "[CV] max_features=auto, n_estimators=20, max_depth=20 ................\n",
      "[CV]  max_features=auto, n_estimators=20, max_depth=20, score=0.133406 - 3.3min\n",
      "[CV] max_features=auto, n_estimators=20, max_depth=20 ................\n",
      "[CV]  max_features=auto, n_estimators=20, max_depth=20, score=0.138348 - 3.2min"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Done   1 jobs       | elapsed:  3.3min\n",
      "[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:  6.5min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "BEST {'max_features': 'auto', 'n_estimators': 20, 'max_depth': 20} 0.13587711428 [mean: 0.13588, std: 0.00247, params: {'max_features': 'auto', 'n_estimators': 20, 'max_depth': 20}] <function _passthrough_scorer at 0x00000000204DB7B8>\n",
      "Best score:  0.13587711428\n",
      "Wall time: 13min 30s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# Define a grid of parameters over which to optimize the random forest\n",
    "# We will figure out which number of trees is optimal\n",
    "parameters = {\"n_estimators\": [20],\n",
    "              \"max_features\": [\"auto\"], # [\"auto\",\"sqrt\",\"log2\"]\n",
    "              \"max_depth\": [20]}\n",
    "best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=2, verbose=3)\n",
    "#best = cv_optimize(estimator, parameters, X_train, y_train, n_folds=2, score_func='mean_squared_error', verbose=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RMSE=  0.111710488577\n"
     ]
    }
   ],
   "source": [
    "#Without weighted sampling: 20 estimators, 20 depth\n",
    "print \"RMSE= \",np.sqrt(mean_squared_error(best.predict(X_test),y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Backup"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Outcome\n",
    "\n",
    "The best RMSE value that we got was 0.120. In NYC each longitude is approx 53 miles & latitude is approx 69 miles (see reference below). This gives an **error range of 6.36 x 8.28 square miles**. So we do not have a great predictor here (but a great learning experience in modeling this problem)\n",
    "\n",
    "note: 1. We have not standardized lat & long in same scale - ideally this shd be done part of data prep. This would have given us a RMSE that we can apply properly to find exact error in distance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}