load_best_model

hcrystalball.model_selection.load_best_model(*, expert_type='best_model', partition_label=None, partition_hash=None, path='')

Returns unpickled file or json metadata from directory.

Parameters

• expert_type (string) – options for json format [‘partition’,’model_reprs’,’logical_partition’,’physical_partition’, ‘frequency’,’horizon’,’best_model_hash’,’best_model_name’,’best_model_repr’] anything else is pickled (e.g. ‘best_model’, ‘cv_data’, ‘cv_results’)

• partition_label (dict) – Model’s metadata used to generate hash for .pkl file. e.g. {‘country’:’Canada’, ‘city’:’Vancouver’, ‘series’:’Number of citizens’}

• partition_hash (string) – Partition’s hash generated with generate_partition_hash function

• path (str) – Path to the directory, where files are to be stored, by default ‘’ resulting in current working directory behaviour

Returns

Return type

unpickled expert