def get_data(symbol, test=None, predictions=None):
history = yfinance.Ticker(symbol).history(period=str(num_years) + 'y',
auto_adjust=False).reset_index()
history = get_ta(history, volume=True, pattern=False)
history.columns = map(str.lower, history.columns)
history['date'] = pd.to_datetime(history['date'])
history['next_return'] = history['close'].pct_change(1).shift(-1)
pos_cutoff = history[history['next_return'] > 0]['next_return'].mean()
neg_cutoff = history[history['next_return'] < 0]['next_return'].mean()
history['return_class'] = 0
#history.loc[ history['next_return']<neg_cutoff, 'return_class'] = -1
#history.loc[ history['next_return']>0, 'return_class'] = 1
history.loc[history['next_return'] > pos_cutoff / 6, 'return_class'] = 1
#history.loc[ history['next_return']>pos_cutoff*2, 'return_class'] = 3
history = history.dropna().reset_index(drop=True)
print('cutoffs %.2f%%' % (pos_cutoff * 100))
print('distribution', history.groupby(by=['return_class'])['open'].count())
print(history.groupby(by=['return_class'])['open'].count() / len(history))
print(history.groupby(by=['return_class'])['next_return'].mean())
return history
def get_data(self, with_decision_tree=False):
all_data = []
for ticker in self.tickers:
ticker_data = yfinance.Ticker(ticker)
ticker_data = ticker_data.history(period=self.period, auto_adjust=False)
ticker_data = ticker_data.reset_index()
ticker_data.columns = map(str.lower, ticker_data.columns)
ticker_data["return"] = ticker_data["close"].pct_change()
#ticker_data["return"] = ticker_data["close"] / ticker_data["open"] - 1
#ticker_data['target'] = ticker_data["close"].shift(-self.hold_length) / ticker_data["close"] - 1
ticker_data['intraday'] = ticker_data["close"] / ticker_data["open"] - 1
ticker_data = ticker_data.drop(columns=['dividends','stock splits'])
ticker_data = get_ta(ticker_data, pattern = True)
ticker_data.columns = map(str.lower, ticker_data.columns)
ticker_data["ticker"] = ticker
ticker_data["range"] = (ticker_data["high"]/ticker_data["low"])-1
ticker_data.dropna(how="any", inplace=True)
all_data.append(ticker_data)
self.history_df = pd.concat(all_data)
self.history_df = self.history_df.sort_values(by=['date'])
#self.history_df = pd.concat([self.all_historic_data, ticker_data])
print(self.history_df)
def get_data(self):
all_historic_data = []
for ticker in self.tickers:
#print('getting data for', ticker)
ticker_data = yfinance.Ticker(ticker)
ticker_data = ticker_data.history(period=self.period, auto_adjust=False)
ticker_data = get_ta(ticker_data, True, self.pattern, self.hold_length)
ticker_data = ticker_data.reset_index()
ticker_data.columns = map(str.lower, ticker_data.columns)
ticker_data["return"] = ticker_data["close"].pct_change()
ticker_data["range"] = (ticker_data["high"]/ticker_data["low"])-1
ticker_data = ticker_data.drop(columns=['dividends','stock splits'])
ticker_data["ticker"] = ticker
ticker_data.dropna(how="any", inplace=True)
ticker_data = ticker_data.reset_index(drop=True)
all_historic_data.append(ticker_data)
self.history_df = pd.concat(all_historic_data)
self.history_df = self.history_df.dropna(thresh=100,axis=1)
self.history_df = self.history_df.sort_values(by=['date'])
self.history_df = self.history_df.reset_index(drop=True)
self.history_df.replace([np.inf, -np.inf], np.nan)
self.history_df = self.history_df.dropna()
self.history_df = self.history_df.replace([np.inf, -np.inf], np.nan)
self.history_df = self.history_df.dropna()
def get_data(symbol, period):
history = yfinance.Ticker(symbol).history(period=period,
auto_adjust=False).reset_index()
history = get_ta(history, volume=True, pattern=False)
history.columns = map(str.lower, history.columns)
history['return'] = history['close'].pct_change(1)
history['next_return'] = history['return'].shift(-1)
history = history.dropna().reset_index(drop=True)
return history
def get_data(self):
stock_df = yfinance.Ticker('QQQ').history(period='7y')
stock_df = stock_df.reset_index()
stock_df = get_ta(stock_df, volume=True, pattern=False)
stock_df.columns = map(str.lower, stock_df.columns)
stock_df["return"] = stock_df["close"].pct_change()
stock_df.columns = map(str.lower, stock_df.columns)
stock_df = stock_df.dropna()
self.clean_train = stock_df[stock_df['date'] < self.cutoff_date]
self.clean_test = stock_df[stock_df['date'] > self.cutoff_date]
def get_data(symbol):
history = yfinance.Ticker(symbol).history(period='7y').reset_index()
history = get_ta(history, volume=True, pattern=False)
history.columns = map(str.lower, history.columns)
history['return'] = history['close'].pct_change() * 100
history = history.dropna()
history['next_return'] = history['return'].shift(-1)
#num_rows = len(history)
#train = history.head( int(num_rows * .75) )
#test = history.tail( int(num_rows *.25) )
history['symbol'] = symbol
return history
def get_data(symbol):
history = yfinance.Ticker(symbol).history(period=str(num_years)+'y', auto_adjust=False).reset_index()
history = get_ta(history, volume=True, pattern=False)
history.columns = map(str.lower, history.columns)
history['date'] = pd.to_datetime(history['date'])
#history['return'] = history['close'].pct_change()
#history['future_price'] = history['close'].shift(-2)
#history['next_return'] = (history['close'].shift(-3) - history['close']) / history['close']
history['next_return'] = history['close'].pct_change(2).shift(-2)
print(history)
#history['next_return'].shift(-2)
pos_cutoff = history[ history['next_return']>0 ]['next_return'].mean()
neg_cutoff = history[ history['next_return']<0 ]['next_return'].mean()
#pos_cutoff = history['return'].mean()
print('pos_cutoff', pos_cutoff, 'neg_cutoff', neg_cutoff)
#print('pos_cutoff', pos_cutoff)
history['return_class'] = 0
history.loc[ history['next_return']>pos_cutoff/2, 'return_class'] = 1
history.loc[ history['next_return']>pos_cutoff, 'return_class'] = 2
history.loc[ history['next_return']>pos_cutoff*2, 'return_class'] = 3
#history['return_class'] = pd.qcut(history['next_return'], q=[0, .33, .66, 1], labels=[0,1,2])
#history.loc[ history['next_return']<neg_cutoff/2, 'return_class'] = -1
#history.loc[ history['next_return']<neg_cutoff, 'return_class'] = -2
history = history.dropna().reset_index(drop=True)
print('distribution',history.groupby(by=['return_class'])['open'].count() / len(history))
print(history.groupby(by=['return_class'])['next_return'].mean())
return history
def get_data(self):
stock_df = yfinance.Ticker('QQQ').history(period='max')
stock_df = stock_df.reset_index()
stock_df = get_ta(stock_df, volume=True, pattern=False)
stock_df.columns = map(str.lower, stock_df.columns)
stock_df["return"] = stock_df["close"].pct_change()
stock_df.columns = map(str.lower, stock_df.columns)
stock_df = stock_df.dropna()
cutoff_datetime = pd.to_datetime(self.cutoff_date)
start_train = cutoff_datetime - timedelta(days=365 * self.train_length)
end_train = cutoff_datetime
start_test = cutoff_datetime
end_test = cutoff_datetime + timedelta(days=365 * 1)
self.clean_train = stock_df[ (stock_df['date'] > start_train) & (stock_df['date'] < end_train) ]
self.clean_test = stock_df[ (stock_df['date'] > start_test) & (stock_df['date'] < end_test) ]
def get_data(tickers, period='5y', pattern=False):
all_historic_data = []
for ticker in tickers:
#print('getting data for', ticker)
ticker_data = yfinance.Ticker(ticker)
ticker_data = ticker_data.history(period=period, auto_adjust=False)
ticker_data = get_ta(ticker_data, True, pattern)
ticker_data = ticker_data.reset_index()
ticker_data.columns = map(str.lower, ticker_data.columns)
ticker_data["return"] = ticker_data["close"].pct_change()
ticker_data["next_day_return"] = ticker_data['close'].shift(
-5) / ticker_data["close"] - 1
ticker_data["range"] = (ticker_data["high"] / ticker_data["low"]) - 1
ticker_data = ticker_data.drop(columns=['dividends', 'stock splits'])
ticker_data["ticker"] = ticker
ticker_data.dropna(how="any", inplace=True)
ticker_data = ticker_data.reset_index(drop=True)
all_historic_data.append(ticker_data)
history_df = pd.concat(all_historic_data)
#print(history_df)
history_df = history_df.dropna(thresh=100, axis=1)
history_df = history_df.replace([np.inf, -np.inf], np.nan)
history_df = history_df.dropna()
history_df = history_df.sort_values(by=['date'])
history_df = history_df.reset_index(drop=True)
return history_df