'''
FIRST SUBMISSION
ACC 0.88864
'''
import sys
import numpy as np
import pandas as pd
import keras
from keras.layers import Dense, Input,Dropout
from keras.models import Sequential
from keras.utils import to_categorical
from keras.callbacks import ModelCheckpoint, TensorBoard, EarlyStopping
import seaborn as sns
import matplotlib.pyplot as plt
import math


def loadCVS(pd_dataName):
    pd_dataName = str(pd_dataName)

    pd_data = pd.read_csv(pd_dataName)

    # fill nan values
    pd_data.replace(np.nan, 0)

    # fill null values
    print(np.sum(pd_data.isnull()))
    pd_data = pd_data.fillna(0)

    # clip from 0
    pd_data['Homepage _Duration'] = pd_data['Homepage _Duration'].clip(lower=0.01)
    pd_data['Homepage _Duration']= np.log(pd_data['Homepage _Duration'])
    pd_data['Aboutus_Duration'] = pd_data['Aboutus_Duration'].clip(lower=0.01)
    pd_data['Aboutus_Duration'] = np.log(pd_data['Aboutus_Duration'])
    pd_data['Contactus_Duration'] = pd_data['Contactus_Duration'].clip(lower=0.01)
    pd_data['Contactus_Duration'] = np.log(pd_data['Contactus_Duration'])


    # remap strings
    month_mapping = {'Jan': 0, 'Feb': 1, 'Mar': 2, 'Apr': 3, 'May': 4, 'Aug': 7, 'Dec': 11, 'Jul': 6, 'June': 5,
                     'Nov': 10, 'Oct': 9, 'Sep': 8}
    pd_data['Month'] = pd_data['Month'].apply(lambda s: month_mapping.get(s) if s in month_mapping else s).astype('int')

    # pd_data['Month']=pd_data['Month'].astype('category')

    vt_mapping = {'Returning_Visitor': 0, 'New_Visitor': 1, 'Other': 2}
    pd_data['VisitorType'] = pd_data['VisitorType'].apply(lambda s: vt_mapping.get(s) if s in vt_mapping else s).astype(
        'int')

    # remap bool
    pd_data['Weekend'] = pd_data['Weekend'].astype('int')

    pd_data = pd_data.astype('float64')

    # print(pd_data.describe())
    print(pd_data.info())

    if 'ID' in pd_data.columns:  # test csv
        Y = None
        pd_data = pd_data.drop('ID', axis=1)
    else:  # train csv
        Y = pd_data['Revenue'].values.reshape((-1, 1))
        Y = to_categorical(Y)
        # pairvise plot
        # sns.pairplot(pd_data,hue='Revenue', vars=['Homepage','Aboutus','Month','TrafficType'])
        # plt.show()

        pd_data = pd_data.drop('Revenue', axis=1)


    X = pd_data.values
    print("X shape before deleting",X.shape)
    xTempMonth=X[:,10]
    np.delete(X,10,axis=0)
    X= (X-np.mean(X))/np.std(X)
    print("X shape after deleting",X.shape)



    xTempMonth=to_categorical(xTempMonth)
    print(X.shape,xTempMonth.shape)
    X=np.concatenate((X,xTempMonth),axis=1)
    print("X shape after concatenating",X.shape)

    # normalize  features
    # pd_data = (pd_data - pd_data.mean()) / pd_data.std()

    # get np array from dataframe

    return X, Y


def nn1():
    mod = Sequential()
    mod.add(Dense(40,activation='sigmoid'))
    # mod.add(Dropout(0.5))
    mod.add(Dense(20,activation='sigmoid'))
    # mod.add(Dropout(0.5))
    mod.add(Dense(2,activation='softmax'))

    return mod


if __name__ == '__main__':
    print("python approach3.py trainSet.csv xTest.csv outpd_data.csv")
    print("python approach3.py [1]          [2]         [3]")
    xTrain, yTrain = loadCVS(sys.argv[1])
    xTest, _ = loadCVS(sys.argv[2])

    print(xTrain.shape, yTrain.shape)

    N=yTrain.shape[0]

    for n in range(N-1,-1,-1):
        # print(n,yTrain[n],yTrain.shape)
        if yTrain[n,0]==1:
            xTrain=np.append(xTrain, np.reshape(xTrain[n],(1,xTrain[n].shape[0])) ,axis=0)
            yTrain=np.append(yTrain, np.reshape(yTrain[n],(1,yTrain[n].shape[0])) ,axis=0)

    a=np.arange(xTrain.shape[0])
    np.random.shuffle(a)
    print(a.shape)
    xTrain=xTrain[a]
    yTrain=yTrain[a]

    print(xTrain.shape, yTrain.shape)
    input("Okay?")

    m = nn1()
    m.compile(loss='binary_crossentropy', optimizer='adam', metrics=['acc'])

    filepath = "weights.hdf5"
    checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode=',min')
    # tensorboard = TensorBoard(log_dir='./logs', write_graph=True, write_grads=False, write_images=False, embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None, embeddings_data=None, update_freq='epoch')
    earlystopping = EarlyStopping(monitor='val_loss', min_delta=1e-3, patience=100, verbose=0, mode='auto',
                                  baseline=None, restore_best_weights=True)

    callbacks_list = [checkpoint, earlystopping]
    hist = m.fit(xTrain, yTrain, validation_split=0.1, epochs=1000, callbacks=callbacks_list,shuffle=True)

    plt.subplot(1, 2, 1)
    plt.plot(hist.history['acc'])
    plt.xlabel('Epoch')
    plt.ylabel("Accuracy")

    plt.subplot(1, 2, 2)
    plt.plot(hist.history['loss'])
    plt.xlabel('Epoch')
    plt.ylabel("Loss")

    plt.show()

    print(m.summary())
    # save model to JSON
    model_json = m.to_json()
    with open("model.json", "w") as json_file:
        json_file.write(model_json)
        print("Model saved to json.")

    # predicting
    yTest = m.predict(xTest)
    yTest = np.argmax(yTest, axis=1)

    pd_test = pd.DataFrame(columns=['ID', 'Revenue'])
    pd_test['ID'] = pd.Series(np.arange(1, len(yTest) + 1))
    pd_test['Revenue'] = pd.Series(yTest)
    pd_test.to_csv(sys.argv[3], index=False)