In [120]:

    

import pandas as pd
from datetime import datetime
import datetime as dt
from pandas_datareader import data, wb
import matplotlib.pyplot as plt
from matplotlib import style


    

In [121]:

    

style.use('ggplot')
def next_biz_day(d):
    nd = d+dt.timedelta(days=1)
    return nd if nd.weekday() in range(5) else next_biz_day(nd)

def prev_biz_day(d):
    pd = d-dt.timedelta(days=1)
    return pd if pd.weekday() in range(5) else prev_biz_day(pd)


    

In [122]:

    

st = prev_biz_day(next_biz_day(datetime(2010,4,5)))
ed = prev_biz_day(next_biz_day(datetime(2017,4,10)))
stock_list = 'BIDU,MSFT,GOOG,AAPL,AMZN,INTC'
target = 'BIDU'
nop = 20
print (st, ed)


    

    




2010-04-05 00:00:00 2017-04-10 00:00:00

In [123]:

    

stocks = stock_list.split(',')
dfs = [data.DataReader(s, 'yahoo', st, ed) for s in stocks] + [
       data.DataReader(s, 'yahoo', st, ed).diff() for s in stocks]
mdfs = []
for i, d in enumerate(dfs):
    d.columns = [c+str(i) for c in d.columns]
    mdfs.append(d)
ds = pd.concat(mdfs, axis=1)


    

In [124]:

    

tst = next_biz_day(st)
ted = next_biz_day(ed)
print(tst, ted)
tdf = data.DataReader(target, 'yahoo', tst, ted).diff()


    

    




2010-04-06 00:00:00 2017-04-11 00:00:00

In [125]:

    

tdf.plot()
plt.show()


    

In [126]:

    

from sklearn import *
lnr = tree.DecisionTreeRegressor()


    

In [127]:

    

print (len(ds),len(tdf))


    

    




1768 1768

In [132]:

    

lnr.fit(ds[2:-nop], tdf['Close'][2:-nop])


    

    
Out[132]:





DecisionTreeRegressor(criterion='mse', max_depth=None, max_features=None,
           max_leaf_nodes=None, min_impurity_split=1e-07,
           min_samples_leaf=1, min_samples_split=2,
           min_weight_fraction_leaf=0.0, presort=False, random_state=None,
           splitter='best')

In [133]:

    

predicted = lnr.predict(ds[-nop:])
out = pd.Series(predicted, tdf[-nop:].index, name='Predicted')
results = (pd.concat([tdf['Close'][-nop:], out], axis=1))


    

In [134]:

    

results.plot()
plt.show()


    

In [135]:

    

results.diff().plot()
plt.show()


    

In [ ]: