Exponential smoothing

polars_ts.models.exponential_smoothing

Pure-Polars exponential smoothing forecasters.

Implements SES, Holt's linear, and Holt-Winters methods. Delegates to Rust when available (2-5x faster per group), falling back to pure Python otherwise. Closes #49.

ses_forecast(df, h, alpha=0.3, target_col='y', id_col='unique_id', time_col='ds')

Forecast with Simple Exponential Smoothing.

Smooths with level parameter alpha and projects a flat forecast.

Parameters

df Input DataFrame. h Forecast horizon. alpha Smoothing parameter for level (0 < alpha < 1).

holt_forecast(df, h, alpha=0.3, beta=0.1, target_col='y', id_col='unique_id', time_col='ds')

Holt's linear trend forecast.

Smooths level and trend, then extrapolates linearly.

Parameters

df Input DataFrame. h Forecast horizon. alpha Smoothing parameter for level. beta Smoothing parameter for trend.

holt_winters_forecast(df, h, season_length, alpha=0.3, beta=0.1, gamma=0.1, seasonal='additive', target_col='y', id_col='unique_id', time_col='ds')

Holt-Winters seasonal forecast.

Smooths level, trend, and seasonal components.

Parameters

df Input DataFrame. h Forecast horizon. season_length Number of observations per season. alpha Smoothing parameter for level. beta Smoothing parameter for trend. gamma Smoothing parameter for seasonality. seasonal "additive" or "multiplicative".