`MultiSampleAnalysis.build_best_model`¶

Module: workflow.py

Evaluates multiple candidate models with CV, computes weighted train/test metrics, and fits best model on full data.

Parameters¶

models (dict[str, Any] | None): Optional model dictionary; defaults include 10 sklearn models.
label_col (str): Label column in feature table.
feature_cols (list[str] | None): Feature columns used for training. Default: ["duration_s", "blockade_ratio", "segment_std", "segment_skew", "segment_kurt"].
cv (int): Number of CV folds (default 10).
scoring (str): Selection metric: accuracy/f1/recall variants.
exclude_noise (bool): Whether to exclude rows marked as noise.

dict model package including best model, selected feature_cols, scores, cv_results, and all_samples_feature_pred (all-sample feature/group table with best_model_pred as last column).