google tends 反指標可行性? FIX_DATA (二)

    def get_historical_interest2(self, keywords, year_start=2018, month_start=1, day_start=1, hour_start=0, year_end=2018, month_end=2, day_end=1, hour_end= 0, cat=0, geo='', gprop='', sleep=0):
        """Gets historical hourly data for interest by chunking requests to 1 week at a time (which is what Google allows)"""

        # construct datetime obejcts - raises ValueError if invalid parameters       
        #start_date = datetime(year_start, month_start, day_start, hour_start)
        #end_date = datetime(year_end, month_end, day_end, hour_end)

        # the timeframe has to be in 1 week intervals or Google will reject it 

        df = pd.DataFrame()

        tf = "all"
        try: 
            self.build_payload(keywords,cat, tf, geo, gprop)
            week_df = self.interest_over_time()
            df = df.append(week_df)
            print (week_df)
        except Exception as e:
            print(e)
            pass

        return df 

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scipy.stats as scs
from matplotlib import animation
from sklearn import preprocessing
from pytrends.request import TrendReq



# Login to Google. Only need to run this once, the rest of requests will use the same session.
pytrend = TrendReq()


tk = pytrend.get_historical_interest2(['stock'], year_start=2018, month_start=8, day_start=1, year_end=2018, month_end=8, day_end=1, cat=0, geo='', gprop='', sleep=0)

#tk.drop(['isPartial'], axis=1)
#del tk['isPartial']
print (tk.head())

plt.figure(figsize=(20,8))
fig, ax = plt.subplots()
#preprocessing.scale(tk['stock'])
ax.plot(tk['stock'])
plt.show()

import numpy as np
import pandas as pd
import csv
import matplotlib.pyplot as plt
import scipy.stats as scs
from matplotlib import animation
from sklearn import preprocessing
from pytrends.request import TrendReq
import datetime

date_time_str = '2018-06-29'  
date_time_obj = datetime.datetime.strptime(date_time_str, '%Y-%m-%d')

df = pd.read_csv('googletrend_stock_data.csv', index_col='date', parse_dates=['date'])
price = df.squeeze()
price.head()


df = pd.read_csv('googletrend_stock_data2.csv', index_col='date', parse_dates=['date'])
all_price = df.squeeze()
all_price.head()
# print('Date:',date_time_str[5:7])  
#Login to Google. Only need to run this once, the rest of requests will use the same session.
# pytrend = TrendReq()


# tk = pytrend.get_historical_interest2(['stock'], year_start=2004, month_start=1, day_start=1, year_end=2018, month_end=8, day_end=1, cat=0, geo='', gprop='', sleep=0)
# print(price)
# #tk.drop(['isPartial'], axis=1)
# #del tk['isPartial']
# print (tk.head())
# type(price)


def get_months( test):
    value = ""
    for i in range (0,len(all_price),1):
        if(test == str(all_price.index[i])[:7]+"-01 00:00:00"):
            value = all_price[str(all_price.index[i])[:7]+"-01 00:00:00"]
    return value

plt.figure(figsize=(16,8))
fig, ax = plt.subplots()

#preprocessing.scale(tk['stock'])
ax.plot(price)
plt.show()


# print( str(price.index[0])[:7] )
# print (price [price.index[0]])
s = pd.Series(price)

fuck = []
print (price.index)


for x in price.index:
    print (x)
    get_month = get_months(str(x)[:7]+"-01 00:00:00" )

    if(type(price[x]) == type(pd.Series(1))):
        fuck.append( [str(x),(np.float64(price[x][0])*(int(get_month)/100))])
    else:
        fuck.append( ([str(x),(np.float64(price[x])*(int(get_month)/100))]))

    #print (get_month/100)
    #print (price[x]*(get_month/100))

    #print (get_months(str(price.index[str(x)])[:7]+"-01 00:00:00"))

df2 = pd.DataFrame(fuck, columns = ['date', 'price'] )
df2['date'] = pd.to_datetime(df2['date'])
price2 = df2.set_index('date')


plt.figure(figsize=(16,8))

fig, ax = plt.subplots()

#preprocessing.scale(tk['stock'])
ax.plot(price2)
plt.gcf().autofmt_xdate()
plt.show()