forecast_forecasting: 3 tools.
| Tool | What it does | Notes |
|---|---|---|
forecast_ml_forecast | Trains machine-learning regressors (LightGBM, XGBoost, CatBoost, random forest, ridge and 40 more) on lag, rolling, calendar and exogenous features of one series (y) or a panel (data: id, time, value) and forecasts h periods ahead, recursively or with one model per step, with conformal prediction intervals. | |
forecast_fitted_values | In-sample predictions of every model over the training history, step periods ahead (1 = one-step), with normal bands from each series’ residual spread at the given levels. | |
forecast_feature_importance | Trains the models on the features and returns what drives them: split/gain importances for tree and boosting models, coefficients for linear models (on standardized features when scale_features), as a table per feature and model and a bar chart of each model’s share of total absolute importance. |
fincept_describe_tool.