# -*- coding: utf-8 -*-
"""
Created on Sat Nov 17 09:03:58 2018

@author: Sri.Venkatesh
"""


import pandas
import os
from sklearn.cross_validation import train_test_split

os.chdir('C:/Users/sri.venkatesh/Desktop/Work/Kaggle/AV/AmEx')

# Given Train Data
Given_Train_Data = pandas.read_csv("./train.csv")
Given_Train_Data['Test_Train_Flag'] = 'Given_Train_Data'
LB_Data = pandas.read_csv("./test.csv")
LB_Data['Test_Train_Flag'] = 'Given_LB_Data'
Entire_Data = pandas.concat([Given_Train_Data,LB_Data],ignore_index = True)
Entire_Data = Entire_Data[Given_Train_Data.columns.tolist()]
user_hist_data = pandas.read_csv('./historical_user_logs.csv')

# Processing Categorical variables
Entire_Data['gender'] = Entire_Data['gender'].fillna('Null')
Entire_Data['user_group_id'] = Entire_Data['user_group_id'].fillna(999)
Entire_Data['user_depth'] = Entire_Data['user_depth'].fillna(999)
Entire_Data['city_development_index'] = Entire_Data['city_development_index'].fillna(999)
product_dummies = pandas.get_dummies(Entire_Data['product'],prefix='product')
campaign_id_dummies = pandas.get_dummies(Entire_Data['campaign_id'],prefix='campaign_id')
user_group_id_dummies = pandas.get_dummies(Entire_Data['user_group_id'],prefix='user_group_id')
user_depth_dummies = pandas.get_dummies(Entire_Data['user_depth'],prefix='user_depth')
gender_dummies = pandas.get_dummies(Entire_Data['gender'],prefix='gender')
city_development_index_dummies = pandas.get_dummies(Entire_Data['city_development_index'],prefix='city_development_index')

# Join the encoded df
Entire_Data = Entire_Data.join(product_dummies)
Entire_Data = Entire_Data.join(campaign_id_dummies)
Entire_Data = Entire_Data.join(user_group_id_dummies)
Entire_Data = Entire_Data.join(user_depth_dummies)
Entire_Data = Entire_Data.join(gender_dummies)
Entire_Data = Entire_Data.join(city_development_index_dummies)
del product_dummies, user_group_id_dummies, user_depth_dummies, gender_dummies, city_development_index_dummies


cols = Entire_Data.columns.tolist()
cols = cols[15:16] + cols[14:15] + cols[:14] + cols[16:]
Entire_Data = Entire_Data[cols]


user_hist_data_agg2 = user_hist_data.pivot_table(values = 'DateTime', index= ['user_id','product'], columns = ['action'], aggfunc = 'size',fill_value = 0).reset_index()
Entire_Data = pandas.merge(Entire_Data,user_hist_data_agg2,on = ['user_id','product'],how = 'left')
Entire_Data['interest'] = Entire_Data['interest'].fillna(0)
Entire_Data['view'] = Entire_Data['view'].fillna(0)
Entire_Data['product_A_view'] = Entire_Data['product_A']*Entire_Data['view'].fillna(0)
Entire_Data['product_B_view'] = Entire_Data['product_B']*Entire_Data['view'].fillna(0)
Entire_Data['product_C_view'] = Entire_Data['product_C']*Entire_Data['view'].fillna(0)
Entire_Data['product_D_view'] = Entire_Data['product_D']*Entire_Data['view'].fillna(0)
Entire_Data['product_E_view'] = Entire_Data['product_E']*Entire_Data['view'].fillna(0)
Entire_Data['product_F_view'] = Entire_Data['product_F']*Entire_Data['view'].fillna(0)
Entire_Data['product_G_view'] = Entire_Data['product_G']*Entire_Data['view'].fillna(0)
Entire_Data['product_H_view'] = Entire_Data['product_H']*Entire_Data['view'].fillna(0)
Entire_Data['product_I_view'] = Entire_Data['product_I']*Entire_Data['view'].fillna(0)
Entire_Data['product_J_view'] = Entire_Data['product_J']*Entire_Data['view'].fillna(0)

Entire_Data['product_A_interest'] = Entire_Data['product_A']*Entire_Data['interest'].fillna(0)
Entire_Data['product_B_interest'] = Entire_Data['product_B']*Entire_Data['interest'].fillna(0)
Entire_Data['product_C_interest'] = Entire_Data['product_C']*Entire_Data['interest'].fillna(0)
Entire_Data['product_D_interest'] = Entire_Data['product_D']*Entire_Data['interest'].fillna(0)
Entire_Data['product_E_interest'] = Entire_Data['product_E']*Entire_Data['interest'].fillna(0)
Entire_Data['product_F_interest'] = Entire_Data['product_F']*Entire_Data['interest'].fillna(0)
Entire_Data['product_G_interest'] = Entire_Data['product_G']*Entire_Data['interest'].fillna(0)
Entire_Data['product_H_interest'] = Entire_Data['product_H']*Entire_Data['interest'].fillna(0)
Entire_Data['product_I_interest'] = Entire_Data['product_I']*Entire_Data['interest'].fillna(0)
Entire_Data['product_J_interest'] = Entire_Data['product_J']*Entire_Data['interest'].fillna(0)

Entire_Data = Entire_Data.drop(['product_J',
                                'campaign_id_414149',
                                'user_group_id_999.0',
                                'user_depth_999.0',
                                'gender_Null',
                                'city_development_index_999.0','view','interest'],axis =1)

#Entire_Data.head(5).to_csv('train_all_cols_sample.csv',index = None)

X = Entire_Data[Entire_Data['Test_Train_Flag'] == 'Given_Train_Data'].iloc[:, 15:]#.values
y = Entire_Data[Entire_Data['Test_Train_Flag'] == 'Given_Train_Data'].iloc[:, 1].values

# Splitting the dataset into the Training set and Test set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.25, random_state = 0)

# Feature Scaling
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

# Fitting Logistic Regression to the Training set
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC
from sklearn.naive_bayes import GaussianNB
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier

LR_classifier = LogisticRegression(random_state = 1)
KNN_classifier = KNeighborsClassifier(n_neighbors = 5, metric = 'minkowski', p = 2)
SVM_classifier = SVC(kernel = 'linear', random_state = 0)
SVMK_classifier = SVC(kernel = 'rbf', random_state = 0)
NB_classifier = GaussianNB()
DT_classifier = DecisionTreeClassifier(criterion = 'entropy', random_state = 0)
RF_classifier = RandomForestClassifier(n_estimators = 500, criterion = 'entropy', random_state = 0)

LR_classifier.fit(X_train, y_train)
#KNN_classifier.fit(X_train, y_train)
#SVM_classifier.fit(X_train, y_train)
#SVMK_classifier.fit(X_train, y_train)
NB_classifier.fit(X_train, y_train)
DT_classifier.fit(X_train, y_train)
RF_classifier.fit(X_train, y_train)

# Predicting the Test set results
y_pred = LR_classifier.predict(X_test)

# Making the Confusion Matrix
from sklearn.metrics import confusion_matrix
cm1 = confusion_matrix(y_test, LR_classifier.predict(X_test))
#cm2 = confusion_matrix(y_test, KNN_classifier.predict(X_test))
#cm3 = confusion_matrix(y_test, SVM_classifier.predict(X_test))
#cm4 = confusion_matrix(y_test, SVMK_classifier.predict(X_test))
cm5 = confusion_matrix(y_test, NB_classifier.predict(X_test))
cm6 = confusion_matrix(y_test, DT_classifier.predict(X_test))
cm7 = confusion_matrix(y_test, RF_classifier.predict(X_test))

from sklearn.metrics import classification_report
print(classification_report(y_test, LR_classifier.predict(X_test)))
#print(roc_auc_score(y_test, KNN_classifier.predict(X_test)))
#print(roc_auc_score(y_test, SVM_classifier.predict(X_test)))
#print(roc_auc_score(y_test, SVMK_classifier.predict(X_test)))
print(classification_report(y_test, NB_classifier.predict(X_test)))
print(classification_report(y_test, DT_classifier.predict(X_test)))
print(classification_report(y_test, RF_classifier.predict(X_test)))

from sklearn.metrics import roc_auc_score
print(roc_auc_score(y_train, LR_classifier.predict(X_train)))
print(roc_auc_score(y_test, LR_classifier.predict(X_test)))
#print(roc_auc_score(y_train, KNN_classifier.predict(X_train)))
#print(roc_auc_score(y_test, KNN_classifier.predict(X_test)))
#print(roc_auc_score(y_train, SVM_classifier.predict(X_train)))
#print(roc_auc_score(y_test, SVM_classifier.predict(X_test)))
#print(roc_auc_score(y_train, SVMK_classifier.predict(X_train)))
#print(roc_auc_score(y_test, SVMK_classifier.predict(X_test)))
print(roc_auc_score(y_train, NB_classifier.predict(X_train)))
print(roc_auc_score(y_test, NB_classifier.predict(X_test)))
print(roc_auc_score(y_train, DT_classifier.predict(X_train)))
print(roc_auc_score(y_test, DT_classifier.predict(X_test)))
print(roc_auc_score(y_train, RF_classifier.predict(X_train)))
print(roc_auc_score(y_test, RF_classifier.predict(X_test)))


from sklearn.ensemble import GradientBoostingClassifier
GB_classifier = GradientBoostingClassifier(random_state = 1)
GB_classifier.fit(X_train, y_train)
cm10 = confusion_matrix(y_test, GB_classifier.predict(X_test))
print(classification_report(y_test, GB_classifier.predict(X_test)))

import xgboost as xgb
xgb_params = {
    'eta': 0.1,
    'max_depth': 10,
    'subsample': 0.6,
    'colsample_bytree': 1,
    'objective': 'binary:logistic',
    'eval_metric': 'auc',
    'silent': 1
}

dtrain = xgb.DMatrix(X_train, y_train)
dtest = xgb.DMatrix(X_test)

#cv_output = xgb.cv(xgb_params, dtrain, num_boost_round=300, early_stopping_rounds=20,
#    verbose_eval=True, show_stdv=False)
#cv_output[['train-rmse-mean', 'test-rmse-mean']].plot()
from sklearn.grid_search import GridSearchCV
from xgboost.sklearn import XGBClassifier
param_test1 = {
 'max_depth':range(3,10,2),
 'min_child_weight':range(1,6,2)
}
gsearch1 = GridSearchCV(estimator = XGBClassifier( learning_rate =0.1, n_estimators=140, max_depth=5,
 min_child_weight=1, gamma=0, subsample=0.8, colsample_bytree=0.8,
 objective= 'binary:logistic', nthread=4, scale_pos_weight=1, seed=27), 
 param_grid = param_test1, scoring='roc_auc',n_jobs=4,iid=False, cv=5)
gsearch1.fit(X_train, y_train)
gsearch1.grid_scores_, gsearch1.best_params_, gsearch1.best_score_

num_boost_rounds = 250
model = xgb.train(dict(xgb_params, silent=1), dtrain, num_boost_round=num_boost_rounds)

y_predict = model.predict(dtest)

#import lightgbm as lgb


# Leader Baord Data Predictions
X_LB = Entire_Data[Entire_Data['Test_Train_Flag'] == 'Given_LB_Data'].iloc[:, 15:].values
X_LB = sc.transform(X_LB)
X_LB = xgb.DMatrix(X_LB)
y_pred_df_lb = pandas.DataFrame(model.predict(X_LB))

LB_Data = pandas.read_csv("./test.csv")
LB_Data['Test_Train_Flag'] = 'Given_LB_Data'
LB_Data = pandas.concat([LB_Data[['session_id']],y_pred_df_lb],axis =1)
LB_Data.to_csv('test15.csv',index = None)