Modeling_v2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>passenger_count</th>\n",
       "      <th>H_Distance</th>\n",
       "      <th>pickup_longitude_range</th>\n",
       "      <th>pickup_latitude_range</th>\n",
       "      <th>dropoff_longitude_range</th>\n",
       "      <th>dropoff_latitude_range</th>\n",
       "      <th>hour</th>\n",
       "      <th>day</th>\n",
       "      <th>month</th>\n",
       "      <th>year</th>\n",
       "      <th>day_of_week</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1.030764</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "      <td>17</td>\n",
       "      <td>15</td>\n",
       "      <td>6</td>\n",
       "      <td>2009</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>8.450134</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>3</td>\n",
       "      <td>6</td>\n",
       "      <td>16</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>2010</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>1.389525</td>\n",
       "      <td>2</td>\n",
       "      <td>6</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>18</td>\n",
       "      <td>8</td>\n",
       "      <td>2011</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>2.799270</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "      <td>21</td>\n",
       "      <td>4</td>\n",
       "      <td>2012</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1.999157</td>\n",
       "      <td>3</td>\n",
       "      <td>6</td>\n",
       "      <td>3</td>\n",
       "      <td>6</td>\n",
       "      <td>7</td>\n",
       "      <td>9</td>\n",
       "      <td>3</td>\n",
       "      <td>2010</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   passenger_count  H_Distance  pickup_longitude_range  pickup_latitude_range  \\\n",
       "0                1    1.030764                       6                      4   \n",
       "1                1    8.450134                       2                      4   \n",
       "2                2    1.389525                       2                      6   \n",
       "3                1    2.799270                       2                      5   \n",
       "4                1    1.999157                       3                      6   \n",
       "\n",
       "   dropoff_longitude_range  dropoff_latitude_range  hour  day  month  year  \\\n",
       "0                        6                       4    17   15      6  2009   \n",
       "1                        3                       6    16    5      1  2010   \n",
       "2                        2                       5     0   18      8  2011   \n",
       "3                        2                       5     4   21      4  2012   \n",
       "4                        3                       6     7    9      3  2010   \n",
       "\n",
       "   day_of_week  \n",
       "0            0  \n",
       "1            1  \n",
       "2            3  \n",
       "3            5  \n",
       "4            1  "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "\n",
    "np.random.seed(42)\n",
    "\n",
    "#Import train and test dataset\n",
    "train = pd.read_csv(\"train.csv\",nrows=10000000)\n",
    "test = pd.read_csv(\"test.csv\")\n",
    "train.tail(10)\n",
    "\n",
    "#Since the missing values is just tiny part of the whole data, we drop them all\n",
    "train = train.dropna()\n",
    "\n",
    "#drop the negative amount since it doesn't make sense.\n",
    "train = train.drop(train[train['fare_amount']<0].index, axis=0)\n",
    "\n",
    "#drop the passenger amount which is larger than 5(each car can take 5 as maximum)\n",
    "train = train.drop(train[train['passenger_count']>5].index, axis = 0)\n",
    "\n",
    "#We can calulate the distance in a sphere when latitudes and longitudes are given by Haversine formula\n",
    "def haversine_distance(lat1, long1, lat2, long2):\n",
    "    data = [train]\n",
    "    for i in data:\n",
    "        R = 6371  #radius of earth in kilometers\n",
    "        #R = 3959 #radius of earth in miles\n",
    "        phi1 = np.radians(i[lat1])\n",
    "        phi2 = np.radians(i[lat2])\n",
    "    \n",
    "        delta_phi = np.radians(i[lat2]-i[lat1])\n",
    "        delta_lambda = np.radians(i[long2]-i[long1])\n",
    "    \n",
    "        #a = sin²((φB - φA)/2) + cos φA . cos φB . sin²((λB - λA)/2)\n",
    "        a = np.sin(delta_phi / 2.0) ** 2 + np.cos(phi1) * np.cos(phi2) * np.sin(delta_lambda / 2.0) ** 2\n",
    "    \n",
    "        #c = 2 * atan2( √a, √(1−a) )\n",
    "        c = 2 * np.arctan2(np.sqrt(a), np.sqrt(1-a))\n",
    "    \n",
    "        #d = R*c\n",
    "        d = (R * c) #in kilometers\n",
    "        i['H_Distance'] = d\n",
    "    return d\n",
    "\n",
    "haversine_distance('pickup_latitude', 'pickup_longitude', 'dropoff_latitude', 'dropoff_longitude')\n",
    "\n",
    "#latitude should between -90 to 90; Longitude should between -180 to 180. pick those are in this range.\n",
    "train = train.drop(train[(train['pickup_latitude']<-90)|(train['pickup_latitude']>90)].index, axis=0)\n",
    "train = train.drop(train[(train['pickup_longitude']<-180)|(train['pickup_longitude']>180)].index, axis=0)\n",
    "train = train.drop(train[(train['dropoff_latitude']<-90)|(train['dropoff_latitude']>90)].index, axis=0)\n",
    "train = train.drop(train[(train['dropoff_longitude']<-180)|(train['dropoff_longitude']>180)].index, axis=0)\n",
    "\n",
    "scale_l = [-74.1,-73.7, 40.55, 40.90]\n",
    "n = 10\n",
    "train = train.drop(train[(train['pickup_latitude']<scale_l[2])|(train['pickup_latitude']>scale_l[3])].index, axis=0)\n",
    "train = train.drop(train[(train['pickup_longitude']<scale_l[0])|(train['pickup_longitude']>scale_l[1])].index, axis=0)\n",
    "train = train.drop(train[(train['dropoff_latitude']<scale_l[2])|(train['dropoff_latitude']>scale_l[3])].index, axis=0)\n",
    "train = train.drop(train[(train['dropoff_longitude']<scale_l[0])|(train['dropoff_longitude']>scale_l[1])].index, axis=0)\n",
    "\n",
    "from sklearn import preprocessing\n",
    "le = preprocessing.LabelEncoder()\n",
    "\n",
    "\n",
    "train['pickup_longitude_range'] = le.fit_transform(pd.cut(train['pickup_longitude'], n))\n",
    "train['pickup_latitude_range'] = le.fit_transform(pd.cut(train['pickup_latitude'], n))\n",
    "train['dropoff_longitude_range'] = le.fit_transform(pd.cut(train['dropoff_longitude'], n))\n",
    "train['dropoff_latitude_range'] = le.fit_transform(pd.cut(train['dropoff_latitude'], n))\n",
    "\n",
    "train = train.drop(['pickup_longitude', 'pickup_latitude', 'dropoff_longitude', 'dropoff_latitude'], axis = 1)\n",
    "\n",
    "train['pickup_datetime'] = pd.to_datetime(arg=train['pickup_datetime'], infer_datetime_format=True)\n",
    "\n",
    "\n",
    "def add_new_date_time_features(dataset):\n",
    "    dataset['hour'] = dataset.pickup_datetime.dt.hour\n",
    "    dataset['day'] = dataset.pickup_datetime.dt.day\n",
    "    dataset['month'] = dataset.pickup_datetime.dt.month\n",
    "    dataset['year'] = dataset.pickup_datetime.dt.year\n",
    "    dataset['day_of_week'] = dataset.pickup_datetime.dt.dayofweek\n",
    "    return dataset\n",
    "\n",
    "train = add_new_date_time_features(train)\n",
    "\n",
    "train.drop(['key'], axis=1, inplace=True)\n",
    "train.drop(['pickup_datetime'], axis=1, inplace=True)\n",
    "\n",
    "\n",
    "\n",
    "y = train.pop('fare_amount')\n",
    "\n",
    "train.head()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LightGBM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/lightgbm/basic.py:1209: UserWarning: categorical_feature in Dataset is overridden.\n",
      "New categorical_feature is ['dropoff_latitude_range', 'dropoff_longitude_range', 'pickup_latitude_range', 'pickup_longitude_range']\n",
      "  'New categorical_feature is {}'.format(sorted(list(categorical_feature))))\n",
      "/usr/local/lib/python3.6/dist-packages/lightgbm/basic.py:762: UserWarning: categorical_feature in param dict is overridden.\n",
      "  warnings.warn('categorical_feature in param dict is overridden.')\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 150 rounds.\n",
      "[50]\tvalid_0's rmse: 6.37109\tvalid_0's l1: 4.03471\n",
      "[100]\tvalid_0's rmse: 4.94873\tvalid_0's l1: 3.02531\n",
      "[150]\tvalid_0's rmse: 4.25856\tvalid_0's l1: 2.47205\n",
      "[200]\tvalid_0's rmse: 3.94253\tvalid_0's l1: 2.18265\n",
      "[250]\tvalid_0's rmse: 3.79463\tvalid_0's l1: 2.02501\n",
      "[300]\tvalid_0's rmse: 3.71891\tvalid_0's l1: 1.94187\n",
      "[350]\tvalid_0's rmse: 3.67496\tvalid_0's l1: 1.89219\n",
      "[400]\tvalid_0's rmse: 3.64744\tvalid_0's l1: 1.86055\n",
      "[450]\tvalid_0's rmse: 3.62729\tvalid_0's l1: 1.83903\n",
      "[500]\tvalid_0's rmse: 3.61178\tvalid_0's l1: 1.82304\n",
      "[550]\tvalid_0's rmse: 3.60034\tvalid_0's l1: 1.81169\n",
      "[600]\tvalid_0's rmse: 3.59116\tvalid_0's l1: 1.80329\n",
      "[650]\tvalid_0's rmse: 3.58378\tvalid_0's l1: 1.79616\n",
      "[700]\tvalid_0's rmse: 3.57722\tvalid_0's l1: 1.7906\n",
      "[750]\tvalid_0's rmse: 3.57131\tvalid_0's l1: 1.78549\n",
      "[800]\tvalid_0's rmse: 3.5654\tvalid_0's l1: 1.78064\n",
      "[850]\tvalid_0's rmse: 3.56062\tvalid_0's l1: 1.77665\n",
      "[900]\tvalid_0's rmse: 3.55619\tvalid_0's l1: 1.7733\n",
      "[950]\tvalid_0's rmse: 3.55227\tvalid_0's l1: 1.77021\n",
      "[1000]\tvalid_0's rmse: 3.54876\tvalid_0's l1: 1.7675\n",
      "[1050]\tvalid_0's rmse: 3.54567\tvalid_0's l1: 1.76503\n",
      "[1100]\tvalid_0's rmse: 3.5429\tvalid_0's l1: 1.76249\n",
      "[1150]\tvalid_0's rmse: 3.54082\tvalid_0's l1: 1.76065\n",
      "[1200]\tvalid_0's rmse: 3.53855\tvalid_0's l1: 1.759\n",
      "[1250]\tvalid_0's rmse: 3.53678\tvalid_0's l1: 1.75755\n",
      "[1300]\tvalid_0's rmse: 3.53459\tvalid_0's l1: 1.75572\n",
      "[1350]\tvalid_0's rmse: 3.53234\tvalid_0's l1: 1.75368\n",
      "[1400]\tvalid_0's rmse: 3.53059\tvalid_0's l1: 1.75228\n",
      "[1450]\tvalid_0's rmse: 3.52895\tvalid_0's l1: 1.75119\n",
      "[1500]\tvalid_0's rmse: 3.5276\tvalid_0's l1: 1.75014\n",
      "[1550]\tvalid_0's rmse: 3.52592\tvalid_0's l1: 1.74902\n",
      "[1600]\tvalid_0's rmse: 3.52454\tvalid_0's l1: 1.7479\n",
      "[1650]\tvalid_0's rmse: 3.5232\tvalid_0's l1: 1.7471\n",
      "[1700]\tvalid_0's rmse: 3.52191\tvalid_0's l1: 1.74613\n",
      "[1750]\tvalid_0's rmse: 3.52075\tvalid_0's l1: 1.7453\n",
      "[1800]\tvalid_0's rmse: 3.51945\tvalid_0's l1: 1.74433\n",
      "[1850]\tvalid_0's rmse: 3.51832\tvalid_0's l1: 1.74362\n",
      "[1900]\tvalid_0's rmse: 3.51742\tvalid_0's l1: 1.74292\n",
      "[1950]\tvalid_0's rmse: 3.51636\tvalid_0's l1: 1.74215\n",
      "[2000]\tvalid_0's rmse: 3.51539\tvalid_0's l1: 1.74145\n",
      "Did not meet early stopping. Best iteration is:\n",
      "[2000]\tvalid_0's rmse: 3.51539\tvalid_0's l1: 1.74145\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "import lightgbm as lgb\n",
    "X_train, X_test, y_train, y_test = train_test_split(train, y, test_size=0.1, random_state=42)\n",
    "lgb_X_train, lgb_X_test, lgb_y_train, lgb_y_test = train_test_split(X_train, y_train, test_size=0.25, random_state=42)\n",
    "lgb_train = lgb.Dataset(lgb_X_train, lgb_y_train)\n",
    "lgb_eval = lgb.Dataset(lgb_X_test, lgb_y_test, reference=lgb_train)\n",
    "# Define hyperparameters\n",
    "params = {\n",
    "    'boosting_type': 'gbdt',\n",
    "    'objective': 'regression',\n",
    "    'num_leaves': 64,\n",
    "    'max_depth': 15,\n",
    "    'min_data_in_leaf': 300,\n",
    "    'learning_rate': 0.01,\n",
    "    'metric': {'l1', 'l2_root'},\n",
    "    'feature_fraction': 0.9,\n",
    "    'bagging_fraction': 0.8,\n",
    "    'bagging_freq': 5,\n",
    "    'verbose': 0\n",
    "}\n",
    "\n",
    "lgbm = lgb.train(params, lgb_train, \n",
    "                 num_boost_round = 2000,\n",
    "                 valid_sets = lgb_eval,\n",
    "                 early_stopping_rounds = 150,\n",
    "                 verbose_eval = 50, \n",
    "                 categorical_feature = ['pickup_longitude_range', \n",
    "                                        'pickup_latitude_range', \n",
    "                                        'dropoff_longitude_range', \n",
    "                                        'dropoff_latitude_range'\n",
    "                                       ]\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The rmse of prediction is: 3.546330741687728\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import mean_squared_error\n",
    "y_pred = lgbm.predict(X_test, num_iteration=lgbm.best_iteration)\n",
    "print('The rmse of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## XGBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/xgboost/core.py:587: FutureWarning: Series.base is deprecated and will be removed in a future version\n",
      "  if getattr(data, 'base', None) is not None and \\\n"
     ]
    }
   ],
   "source": [
    "import xgboost as xgb\n",
    "xgb_train = xgb.DMatrix(X_train, label=y_train)\n",
    "xgb_test = xgb.DMatrix(X_test, label=y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\txgb_train-rmse:11.471\txgb_test-rmse:11.4802\n",
      "Multiple eval metrics have been passed: 'xgb_test-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until xgb_test-rmse hasn't improved in 50 rounds.\n",
      "[100]\txgb_train-rmse:3.5834\txgb_test-rmse:3.60189\n",
      "[200]\txgb_train-rmse:3.53194\txgb_test-rmse:3.57124\n",
      "[300]\txgb_train-rmse:3.50574\txgb_test-rmse:3.56274\n",
      "[400]\txgb_train-rmse:3.48689\txgb_test-rmse:3.55798\n",
      "[500]\txgb_train-rmse:3.47071\txgb_test-rmse:3.5552\n",
      "[600]\txgb_train-rmse:3.45801\txgb_test-rmse:3.55366\n",
      "[700]\txgb_train-rmse:3.44586\txgb_test-rmse:3.55211\n",
      "[800]\txgb_train-rmse:3.4357\txgb_test-rmse:3.55131\n",
      "Stopping. Best iteration:\n",
      "[754]\txgb_train-rmse:3.44021\txgb_test-rmse:3.55085\n",
      "\n"
     ]
    }
   ],
   "source": [
    "params={'objective':'reg:linear',\n",
    "        'eval_metric':'rmse',\n",
    "        'learning_rate': 0.2,\n",
    "        'max_depth': 6,\n",
    "        'max_leaves': 0,\n",
    "        'n_estimators': 100,\n",
    "        'min_child_weight': 1,\n",
    "        'gamma': 0,\n",
    "        'grow_policy': 'lossguide',\n",
    "        'colsample_bytree': 0.85,\n",
    "        'min_child_weight': 1,\n",
    "        'tree_method': 'gpu_hist'}\n",
    "\n",
    "watchlist = [(xgb_train, 'xgb_train'), (xgb_test, 'xgb_test')]\n",
    "xgbm = xgb.train(params = params,\n",
    "                 dtrain = xgb_train,\n",
    "                 evals = watchlist,\n",
    "                 num_boost_round = 2000,\n",
    "                 early_stopping_rounds = 50,\n",
    "                 verbose_eval = 100) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The rmse of prediction is: 3.554639878635857\n"
     ]
    }
   ],
   "source": [
    "y_pred = xgbm.predict(xgb_test)\n",
    "print('The rmse of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fe76f67d358>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "xgb.plot_importance(xgbm, ax=None, height=0.2, xlim=None, ylim=None, title='Feature importance', xlabel='F score', ylabel='Features', importance_type='weight', max_num_features=None, grid=True, show_values=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Linear Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training score: 0.77140\n",
      "Testing score: 0.77119\n",
      "The RMSE of prediction is: 4.357027676129611\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "linreg = LinearRegression(n_jobs=-1)\n",
    "linreg.fit(X_train, y_train)\n",
    "\n",
    "print(\"Training score: {:.5f}\".format(linreg.score(X_train, y_train)))\n",
    "print(\"Testing score: {:.5f}\".format(linreg.score(X_test, y_test)))\n",
    "\n",
    "y_pred = linreg.predict(X_test)\n",
    "\n",
    "print('The RMSE of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ridge Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set score: 0.77140\n",
      "Testing set score: 0.77119\n",
      "The RMSE of prediction is: 4.357027683901787\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import RidgeCV\n",
    "\n",
    "reg_cv = RidgeCV(alphas = [0.001, 0.01, 0.1, 1, 5, 10], cv=5)\n",
    "\n",
    "reg_cv.fit(X_train, y_train)\n",
    "\n",
    "print(\"Training set score: {:.5f}\".format(reg_cv.score(X_train, y_train)))\n",
    "print(\"Testing set score: {:.5f}\".format(reg_cv.score(X_test, y_test)))\n",
    "\n",
    "y_pred = reg_cv.predict(X_test)\n",
    "\n",
    "print('The RMSE of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LASSO"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set score: 0.77140\n",
      "Testing set score: 0.77119\n",
      "The RMSE of prediction is: 4.3570384304403\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LassoCV\n",
    "\n",
    "lasso_cv = LassoCV(alphas = [0.001, 0.01, 0.1, 1, 5, 10], cv=5)\n",
    "\n",
    "lasso_cv.fit(X_train, y_train)\n",
    "\n",
    "print(\"Training set score: {:.5f}\".format(lasso_cv.score(X_train, y_train)))\n",
    "print(\"Testing set score: {:.5f}\".format(lasso_cv.score(X_test, y_test)))\n",
    "\n",
    "y_pred = lasso_cv.predict(X_test)\n",
    "\n",
    "print('The RMSE of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/sklearn/ensemble/forest.py:246: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set score: 0.96677\n",
      "Testing set score: 0.81927\n",
      "The RMSE of prediction is: 3.8722403145679527\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestRegressor\n",
    "\n",
    "rfreg = RandomForestRegressor(n_jobs=-1)\n",
    "# n_estimators = 10\n",
    "# min_samples_split: 2\n",
    "# min_samples_leaf: 1\n",
    "# boostrap: True\n",
    "\n",
    "rfreg.fit(X_train, y_train)\n",
    "\n",
    "print(\"Training set score: {:.5f}\".format(rfreg.score(X_train, y_train)))\n",
    "print(\"Testing set score: {:.5f}\".format(rfreg.score(X_test, y_test)))\n",
    "\n",
    "y_pred = rfreg.predict(X_test)\n",
    "\n",
    "print('The RMSE of prediction is:', mean_squared_error(y_test, y_pred) ** 0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fe76ec9da58>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Feature Importance\n",
    "feat_importances = pd.Series(rfreg.feature_importances_, index=train.columns)\n",
    "feat_importances.nlargest(5).plot(kind='barh')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}