%matplotlib inline

import numpy as np
import pandas as pd
import seaborn as sns

from scipy import stats
from zipline.data import load_from_yahoo

stocks = ['XLY', 'XLF', 'XLK', 'XLE', 'XLV', 'XLI', 'XLP', 'XLB', 'XLU', 'SPY']

end_date = pd.datetools.datetime(2014, 5, 1)

# Ensure there is enough data (number of days: 800)
start_date = end_date - pd.DateOffset(n=800)

data = load_from_yahoo(stocks=stocks, start=start_date, end=end_date)

data.plot()

print data.index[0]
print data.index[-1]

prices = data.as_matrix()[:400]

prices.shape

changes_all = stats.zscore(prices, axis=0, ddof=1)

changes_all.shape # Quick look at the shape

changes = changes_all[:,0:-1] - np.tile(changes_all[:,-1], (len(stocks)-1,1)).T

changes = changes > 0
changes = changes.astype(np.int)

changes.shape

X = np.split(changes[:,0], 20)
X

changes[:,0]

Y = X[-1]
Y

full_dates = data.index[:400]
dates = []

for date in full_dates:
    new_date = str(date)[:10]
    dates.append(new_date)

dates = np.array(dates)

dates[:10]

dates_split = np.array(np.split(dates, 20))

dates_split.shape

dates_split[:2]

dates_labels = dates_split[-1]
dates_labels

for x_dates, label in zip(dates_split, dates_labels):
    print 'Target:', label
    print 'Training:', x_dates
    print ''