In [3]:

    

import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestRegressor
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.metrics import r2_score, mean_squared_error
from sklearn.cross_validation import KFold


    

In [4]:

    

data_path = '/devdata/course_proj/aml/CitiBot/data/'


    

In [5]:

    

train = pd.read_csv(data_path+'train.csv').ix[:,1:]
testSet = pd.read_csv(data_path+'test.csv').ix[:,1:]
# encode features for random forest and concatenate them back to dataframe
train = pd.concat([train, train['season'].str.get_dummies(sep=',')], axis=1)
train = pd.concat([train, train['dayOfWeek'].str.get_dummies(sep=',')], axis=1)

testSet = pd.concat([testSet, testSet['season'].str.get_dummies(sep=',')], axis=1)
testSet = pd.concat([testSet, testSet['dayOfWeek'].str.get_dummies(sep=',')], axis=1)

# we don't have 'spring', 'summer', 'fall' in current test set
cols = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday', 
        'holiday','winter',
        'max', 'min', 'rain' ,'snow']

X_train = train[cols]
y_train = train.visited
X_test = testSet[cols]
y_test = testSet.visited


    

In [6]:

    

#gredient boosting regressor For Feature Extraction

def gbm_features(traned_gbm, data_x):
    # gredient boosting regressor with least square loss function   
    X_gbm = gbm.transform(data_x)
    return X_gbm


    

In [14]:

    

params = {'n_estimators': 200, 'max_depth': 4, 'min_samples_split': 1,
          'learning_rate': 0.01, 'loss': 'ls'}

gbm = GradientBoostingRegressor(**params)
gbm.fit(X_train.values, y_train.values)


    

    
Out[14]:





GradientBoostingRegressor(alpha=0.9, init=None, learning_rate=0.01, loss='ls',
             max_depth=4, max_features=None, max_leaf_nodes=None,
             min_samples_leaf=1, min_samples_split=1,
             min_weight_fraction_leaf=0.0, n_estimators=200,
             random_state=None, subsample=1.0, verbose=0, warm_start=False)

In [17]:

    

X_gbm = gbm.transform(X_test.values)


    

In [19]:

    




    

    
Out[19]:





array([[  4.20000000e+01,   3.40000000e+01,   0.00000000e+00],
       [  4.20000000e+01,   3.40000000e+01,   0.00000000e+00],
       [  4.20000000e+01,   3.40000000e+01,   0.00000000e+00],
       ..., 
       [  6.10000000e+01,   4.70000000e+01,   5.00000000e-02],
       [  6.10000000e+01,   4.70000000e+01,   5.00000000e-02],
       [  6.10000000e+01,   4.70000000e+01,   5.00000000e-02]])