environmentaltools.spatiotemporal.bme.perform_cross_validation

environmentaltools.spatiotemporal.bme.perform_cross_validation(dfh, dfs, zh, covmodel, covparam, nmax, dmax, order, option, path, name, k)

Perform k-fold cross-validation for BME model evaluation.

Assesses BME model performance by splitting hard data into training and validation sets.

Parameters:

• dfh (pd.DataFrame) – Hard data with columns [‘x’, ‘y’, ‘t’, ‘h’].

• dfs (pd.DataFrame) – Soft data with columns [‘x’, ‘y’, ‘t’, ‘h’, ‘s’].

• zh (np.ndarray) – Smoothed hard data values.

• covmodel (str) – Covariance model type.

• covparam (array-like) – Covariance model parameters.

• nmax (list of int) – Maximum neighbors [hard_max, soft_max].

• dmax (list of float) – Maximum distances [spatial_max, temporal_max, space_time_ratio].

• order (list of int) – Regression orders [spatial_order, temporal_order].

• option (list) – BME options [max_intervals, num_moments, percentile].

• path (str) – Directory for caching results.

• name (str) – Base filename for results.

• k (int) – Number of cross-validation folds.

Returns:

• e_mda (list) – Mean absolute errors for each fold.

• e_mse (list) – Root mean squared errors for each fold.

Notes

Results are cached to disk. Uses random subsampling without replacement for each fold.