Pelt

polars_ts.changepoint.pelt

PELT (Pruned Exact Linear Time) changepoint detection.

Delegates to the Rust implementation when available (10-50x faster), falling back to pure Python otherwise.

_cost_mean(data, start, end)

Cost of segment [start, end) under a change-in-mean model.

_cost_var(data, start, end)

Cost of segment [start, end) under a change-in-variance model.

_cost_meanvar(data, start, end)

Cost of segment under a change-in-mean-and-variance model.

_pelt_python(df, target_col, id_col, time_col, cost, penalty, min_size)

Pure-Python PELT implementation (fallback).

_pelt_rust(df, target_col, id_col, time_col, cost, penalty, min_size)

Rust-accelerated PELT implementation.

pelt(df, target_col='y', id_col='unique_id', time_col='ds', cost='mean', penalty=None, min_size=2)

Detect multiple changepoints using the PELT algorithm.

Parameters

df Input DataFrame. target_col Column to analyze. id_col Column identifying each time series. time_col Column with timestamps. cost Cost function: "mean", "var", or "meanvar". penalty Penalty per changepoint. Defaults to 2 * log(n) (BIC-like). min_size Minimum segment length between changepoints.

Returns

pl.DataFrame DataFrame with columns [id_col, "changepoint_idx", time_col] listing detected changepoint locations.