# -*- coding: utf-8 -*-

import pandas as pd

import numpy as np

from numpy import concatenate

from math import sqrt

import pickle

from sklearn.preprocessing import MinMaxScaler

from minepy import MINE

import pandas as pd

from pandas import Series

from pandas import DataFrame

from matplotlib import pyplot

import statsmodels.api as sm

def stl(data):

series = data[['date','arrival']]

series.date = pd.to_datetime(series.date,errors='coerce')

series = series.set_index('date')

res = sm.tsa.seasonal_decompose(series)

residual = res.resid.dropna()

trend = res.trend.dropna()

seaonality = res.seasonal.dropna()

return to_drop

def Scaler(data):

values = data.values

values = values.astype('float32')

scaler = MinMaxScaler(feature_range=(0,1))

scaled = scaler.fit_transform(values)

scaledDf = pd.DataFrame(scaled,columns=data.columns)

return scaler,scaledDf

def series_to_supervised(data):

dataset1 = data

# manually specify column names

dataset1.index.name = 'date'

df_finalfeature = dataset1.iloc[:,4:]

df_finalfeature.replace(np.nan, 0, inplace=True)

np.where(np.isnan(df_finalfeature.values))

scaler = sklearn.preprocessing.MinMaxScaler()

df_finalfeature__nor = pd.DataFrame(scaler.fit_transform(df_finalfeature), columns=df_finalfeature.columns)

# x__nor = x

df_finalfeature__nor.describe()

df_df = pd.concat([df_finalfeature__nor,dataset1['arrival']/1000000,dataset1['seasonality']/1000000,dataset1['trend']/1000000],axis = 1)

df_df = df_df[6:86]

df_1 = df_df.as_matrix() # convert to numpy array

data_1 = []

# create all possible sequences of length seq_len

for index in range(len(df_1) - 13):

data_1.append(df_1[index: index + 13])

data_1 = np.array(data_1)

y_1 = np.zeros((len(data_1), 1))

for i in range(0,len(y_1)):

y_1[i]=data_1[i][12][77]

x_1 = np.zeros((len(data_1),12,78))

for i in range(0,len(x_1)):

x_1[i]=data_1[i][0:12]

return x_1,y_1