adaptive_volatility

Defined in fynance.features.engineering

adaptive_volatility(X, windows, regimes, period=252)

Regime-adaptive realized volatility (worked example of adaptive_roll).

Uses a short volatility window in some regimes and a longer one in others, so the estimate reacts fast in turbulent regimes and stays smooth in calm ones.

Parameters:

Xnp.ndarray

One-dimensional price/level series.

windowsmapping of int to int

Volatility window for each regime label.

regimesnp.ndarray

Causal integer regime label per bar, aligned with X.

periodint, optional

Annualization factor. Default 252.

Returns:

np.ndarray

Regime-adaptive annualized volatility, shape (len(X),).

Examples

>>> import numpy as np
>>> rng = np.random.default_rng(0)
>>> X = 100 * np.exp(np.cumsum(rng.standard_normal(100) * 0.01))
>>> regimes = (np.arange(100) // 50)   # two regimes
>>> adaptive_volatility(X, {0: 5, 1: 20}, regimes).shape
(100,)