Skip to main content
def Agent4PredictorWorker.fit(self, inputs, fx_choice=REG.LGB, **params) [source]
Train the GBM agent on numerical and/or categorical features. Executes the full two-stage feature selection pipeline (tree-based selection + shadow pruning), fits the final GBM with early stopping, and computes validation diagnostics.

​
Parameters

inputs : Agent4Input
Training configuration including train_set,
num_features, and cat_features.
 fx_choice : str
The GBM engine: REG.LGB (LightGBM, default)
or REG.XGB (XGBoost).
 **params : dict
Runtime parameters passed to the internal fitter
(gb_max_leaves, gb_learning_rate, gb_estimators,
gb_min_child_weight, shadow_quantile, cross_val_folds).

​
Returns

Agent4Predictor
The fitted agent instance.