import requests

import json

import csv

import os

import time

import numpy as np

import matplotlib.pyplot as plt

from sklearn.externals import joblib

from mlhelper import readCsvIntoPandasDataframe, preprocessData, preprocessDataWithoutScale, plot_learning_curve

from sklearn import neighbors, cross_validation, metrics, ensemble, svm

from sklearn.cross_validation import KFold, StratifiedKFold

def plotImportances(clf, header, title):

return header[sorted_idx], feature_importance[sorted_idx]

csvfile = 'HousePriceList_All_Train.csv'

# adding attributes did not make a difference

#csvfile = 'rentList_All_withAllAttr_trimmed.csv'

#csvfile = 'rentList_E_EC_N_final.csv'

data = readCsvIntoPandasDataframe(csvfile)

X, y, X_scaler, y_scaler, header, imp, vec = preprocessData(data)

#print "header type:"

#print type(header).__name__

#print header

#X, y = preprocessDataWithoutScale(data)

joblib.dump(X_scaler, 'pickle-house/X_scaler.pkl')

joblib.dump(y_scaler, 'pickle-house/y_scaler.pkl')

joblib.dump(imp, 'pickle-house/Imputer.pkl')

joblib.dump(vec, 'pickle-house/Vector.pkl')

estimators = []

# K-Nearest Neighbors

estimators.append( {"name": "KNN",

"model": neighbors.KNeighborsRegressor(

weights = "uniform", n_neighbors = 5) } )

# Gradient Boosting Regressor

estimators.append( {"name": "GBR",

"model": ensemble.GradientBoostingRegressor(

max_features = 0.1, n_estimators = 2100,

max_depth = 6, min_samples_leaf = 1, learning_rate = 0.02) })

# Random Forest

estimators.append( {"name": "RF",

"model": ensemble.RandomForestRegressor(

max_features = 0.1, n_estimators = 512 ) })

# Support Vector Machine

estimators.append( {"name": "SVR",

"model": svm.SVR(cache_size=1000, kernel='poly',

C = 1, gamma = 0.1, epsilon = 0.1, degree = 3) })

#kf = KFold(len(X), n_folds=10, shuffle=True, random_state=17) #3) #70) #48)

#kf = StratifiedKFold(y, n_folds=10, shuffle=True, random_state=3) #70) #48)

for estimator in estimators:

print "== " + estimator['name'] + " =="

model = estimator['model']

start = time.time()

model.fit(X, y)

print("Fitting took %.2f seconds." % (time.time() - start) )

joblib.dump(model, 'pickle-house/' + estimator['name'] + '.pkl')

if(estimator['name'] == 'GBR' or estimator['name'] == 'RF'):

title = estimator['name']

hdr, imp = plotImportances(model, header, title)

#print hdr

#print imp

#o = np.concatenate((hdr, imp), axis=0)

o = np.column_stack((hdr, imp))

np.savetxt("pickle-house/FeatImp-" + estimator['name'] + ".csv",

o, delimiter=",", fmt="%s")

#fout = open("pickle/FeatImp-" + estimator['name'] + "-" + str(i) + ".csv","a")

#fout.write(hdr)

#fout.write(imp)

#fout.close()