Module decomposition 

Seasonal-Trend decomposition using Loess (STL).

Splits a time series into trend, seasonal, and residual components: y = trend + seasonal + residual.

The implementation follows the Cleveland et al. (1990) STL algorithm:

1. Start with seasonal = 0.
2. Detrend: detrended = y - trend.
3. Subseries smoothing: extract each season’s slice, apply degree-1 Loess (locally weighted polynomial regression with tricube kernel).
4. Trend update: y - seasonal, then Loess with trend_window.
5. Repeat for n_iter iterations.

Structs

Decomposition

Decomposition result: y[i] = trend[i] + seasonal[i] + residual[i].

StlConfig

Configuration for the STL decomposer.

StlDecomposer

STL decomposer implementing SeasonalDecomposer.

Enums

DecompositionError

Errors that can occur during decomposition.

Traits

SeasonalDecomposer

Trait for seasonal decomposition strategies.

Functions

detect_period

Detect the dominant period using autocorrelation peak finding.

detect_period_with_threshold

Detect the dominant period using autocorrelation peak finding with a caller-specified minimum ACF threshold.

stl_decompose

Perform STL decomposition.