2023.4.23 修正

ウィンドウサイズの決定と減退係数の決定

取引量のボラに基づいて算出

code: python

log_returns_volume = np.log(df'volume' / df'volume'.shift(1))

volume_volatility = log_returns_volume.std()

if volume_volatility > high_volatility_threshold:

window_size = short_window_size

elif volume_volatility < low_volatility_threshold:

window_size = long_window_size

else:

# 中間レベルのボラティリティの場合、適切なウィンドウサイズを選択

window_size = intermediate_window_size

alpha = 2 / (window_size + 1)

df'ema' = df'price'.ewm(alpha=alpha).mean()

減退係数α

ボラが高い場合は短いウィンドウ、ボラが低い場合は長いウィンドウを採用

上記の方法に、過去データと最新データの2つを組み合わせてバランスをとるParcl方式を融合

code:python

log_returns_volume = np.log(df'volume' / df'volume'.shift(1))

volume_volatility = log_returns_volume.std()

if volume_volatility > high_volatility_threshold:

window_size = 7

elif volume_volatility < low_volatility_threshold:

window_size = 30

else:

# 中間レベルのボラティリティの場合、適切なウィンドウサイズを選択（線形補間を利用）

window_size = 7 + (30 - 7) * ((volume_volatility high_volatility_threshold) / (low_volatility_threshold - high_volatility_threshold))**2

alpha = 2 / (window_size + 1)

W_ti = 1/alpha

W_si = 1 - W_ti

Index = (ps_i * w_si) + (pt_i * w_ti)