Pipeline API

ForecastPipeline

ForecastPipeline(steps)

Sequential pipeline that chains transformers with a forecaster.

Methods

• fit(y) → self
• predict(steps) → (predictions, lower, upper)
• transform(y) → transformed data
• inverseTransform(y) → original scale data
• listSteps() → list of step names
• getStep(name) → transformer/forecaster instance
• getParams() → dict of all parameters

Transformers

All transformers implement fitTransform(y) and inverseTransform(y).

Scaler

Scaler(method='zscore')

Methods: 'zscore' (mean=0, std=1), 'minmax' (0-1 range)

LogTransformer

LogTransformer()

log(1 + y) with automatic shift for negative values.

BoxCoxTransformer

BoxCoxTransformer(lmbda=None)

Optimal Box-Cox lambda via MLE. Pass lmbda to fix.

Differencer

Differencer(d=1)

d-th order differencing.

Deseasonalizer

Deseasonalizer(period=7)

Remove seasonal component by period averaging.

Detrend

Detrend()

Remove linear trend.

OutlierClipper

OutlierClipper(factor=3.0)

IQR-based outlier clipping. No inverse transform.

MissingValueImputer

MissingValueImputer(method='linear')

Methods: 'linear', 'mean', 'ffill'. No inverse transform.