def forward(self, tensor_price, tensor_volume):
rocp = tsf.pct_change(tensor_price)
tensor_volume = torch.clamp(tensor_volume, min=1)
v_pct_change = tsf.pct_change(tensor_volume)
v_pct_change_atan = torch.atan(v_pct_change)
pv = rocp * v_pct_change_atan
return pv
def forward(self, tensor):
"""defalt input is close"""
returns = tsf.pct_change(tensor, period=1)
sharp_ratio = tsf.rolling_mean_(returns,
window=self._window) / tsf.rolling_std(
returns, window=self._window)
return tsf.shift(sharp_ratio, window=self._lag_window).squeeze(-1)
def forward(self, tensor):
tensor_rolling_mean = tsf.rolling_mean_(
tensor, window=self.timeperiod).squeeze(-1)
volume_pct_change = tsf.pct_change(tensor_rolling_mean)
return torch.atan(volume_pct_change)
def forward(self, tensor):
tensor_rolling_mean = tsf.rolling_mean_(
tensor, window=self.timeperiod).squeeze(-1)
return tsf.pct_change(tensor_rolling_mean)
def forward(self, tensor):
tensor = torch.clamp(tensor, min=1)
pct_change = tsf.pct_change(tensor)
return torch.atan(pct_change)
def forward(self, open, close):
diff = close - open
output = diff.div(open)
output_ts = tsf.pct_change(output, period=self._window)
return output_ts
def forward(self, open, close):
shift_tensor = tsf.shift(close, window=1)
diff = open - shift_tensor
output = diff.div(shift_tensor)
output_ts = tsf.pct_change(output, period=self._window)
return output_ts
def forward(self, tensor):
return tsf.pct_change(tensor)
def forward(self, tensor):
rsi_ts = talib_func.rsi(tensor, timeperiod=self.timeperiod).squeeze(-1)
rsi_pct_change = tsf.pct_change(rsi_ts + 100)
return rsi_pct_change
def forward(self, tensor):
return tsf.pct_change(tensor, period=self._window)
import featurizer.functors.talib as talib
import featurizer.functions.talib_functions as talib_func
import torch
import numpy as np
import pandas as pd
# fake data
# type=tensor, shape=50x10
open = torch.tensor(np.random.uniform(low=20, high=40, size=(50, 10)))
close = torch.tensor(np.random.uniform(low=20, high=40, size=(50, 10)))
high = torch.tensor(np.random.uniform(low=30, high=50, size=(50, 10)))
low = torch.tensor(np.random.uniform(low=10, high=30, size=(50, 10)))
volume = torch.tensor(np.random.uniform(low=10000, high=50000, size=(50, 10)))
total_turnover = torch.tensor(
np.random.uniform(low=200000, high=2000000, size=(50, 10)))
returns = tsf.pct_change(close, 1)
# ROCP
func_ROCP = talib.ROCP(timeperiod=1)
feature_ROCP = func_ROCP(close)
# VolumeROCP
func_VROCP = talib.VolumeROCP()
feature_VROCP = func_ROCP(volume)
# MAROCP
func_MAROCP = talib.MAROCP(timeperiod=5)
feature_MAROCP = func_MAROCP(close)
# MARelative
func_MA = talib.MARelative(timeperiod=5)
def forward(self, tensor):
returns = tsf.pct_change(tensor, period=self._window)
return tsf.shift(returns, window=-self._window)
def forward(self, tensor):
"""input is close"""
tensor = tsf.pct_change(tensor, period=self._window)
return tsf.shift(tensor, window=self._lag_window)
