Baseline Forecasting

Builds and evaluates simple baseline forecasts with ForecasterEquivalentDate (equivalent-date / seasonal-naive / moving-average) to benchmark machine learning models. Covers offset (int vs pandas DateOffset), n_offsets with agg_func, backtesting a baseline, comparing against ML with MASE, conformal intervals, and grid_search_equivalent_date. Use when the user wants a baseline, a naive or seasonal reference, or to check whether a model beats a simple rule. Not for choosing among ML forecasters (see choosing-a-forecaster) or tuning ML hyperparameters (see hyperparameter-optimization).

skforecast Updated 1.5k repo stars
skforecast

File contents

skforecast/skforecast commit 151679b781

Frequently asked questions

npx get-skill add skforecast/baseline-forecasting