anofox-forecast

Run `npx skills add DataZooDE/anofox-forecast --all -g -y` in your terminal to install all skills in this suite globally.

Load the extension and call ts_forecast_by with your table, date column, target column, a model name like AutoETS, and a horizon. It returns point forecasts with confidence intervals as a regular SQL table.

It includes 36 models such as AutoETS, AutoARIMA, Theta, Holt-Winters, GARCH, Kalman filters, Croston for intermittent demand, and global panel models for many related series.

Yes. Use ts_cv_folds_by to create time-aware cross-validation folds, ts_cv_forecast_by to forecast each fold, and scalar metrics like ts_mae or ts_rmse with GROUP BY to compare models.

No. Everything runs as SQL functions inside DuckDB, so anyone who can write a query can forecast, detect seasonality, and score data quality without Python or R.