API reference

Everything below is importable from the top-level pymc_forecast namespace; the modules group the reference by responsibility.

• pymc_forecast.model — Model-building core: Horizon, time_series, predict, build_model, and the ForecastingModel facade.

• pymc_forecast.forecaster — Forecaster classes: Forecaster (ADVI), HMCForecaster (NUTS), PathfinderForecaster.

• pymc_forecast.prediction — Posterior-predictive plumbing: forecast, predict_in_sample, thin_draws.

• pymc_forecast.evaluate — Rolling/expanding-window backtesting: backtest, BacktestResult, results_to_dataframe.

• pymc_forecast.metrics — Dim-aware forecast metrics: CRPS, pinball, interval score, coverage, MASE, evaluate_forecast.

• pymc_forecast.features — Feature builders: Fourier design matrices and periodic tiling.

• pymc_forecast.datasets — Example datasets: BART ridership and Victoria electricity demand.

• pymc_forecast.data — Labeled-array helpers: normalization to DataArray, time-index extension, alignment checks.

• pymc_forecast.priors — pymc-extras Prior interop: declarative, user-injectable priors for the model primitives.

• pymc_forecast.gaussian — Time-correlated Gaussian observation noise: predict_mvn and the explicit Gaussian conditional.

• pymc_forecast.markov — Scan-based Markov latents: markov_time_series.

• pymc_forecast.statespace — pymc-extras statespace interop: StatespaceModel and StatespaceForecaster.

• pymc_forecast.exceptions — The package’s error taxonomy.