Module features 

Temporal feature extraction functions for time-series data.

All functions operate on &[f64] slices and return Vec<f64>, using f64::NAN where values cannot be computed (e.g. boundary rows).

Functions

diff

N-th order differencing.

ewm

Exponentially weighted mean with smoothing factor alpha ∈ (0, 1].

lag

Lag: value offset rows earlier. Returns NAN for first offset rows.

pct_change

Percentage change: (values[i] - values[i-1]) / values[i-1]. First element is NAN. Returns NAN when values[i-1] is zero.

rolling_corr

Rolling Pearson correlation over a trailing window, pairwise-complete.

rolling_mean

Rolling arithmetic mean of the non-NULL values in a trailing window.

rolling_std

Rolling sample standard deviation of the non-NULL values in a trailing window, by Welford’s algorithm.

zscore

Z-score normalisation against the partition’s own mean and sample standard deviation.