dabl.AnyClassifier¶
- class dabl.AnyClassifier(n_jobs=None, min_resources='exhaust', verbose=0, type_hints=None, portfolio='baseline')[source]¶
Classifier with automatic model selection.
This model uses successive halving on a portfolio of complex models (HistGradientBoosting, RandomForest, SVC, LogisticRegression) to pick the best model family and hyper-parameters.
AnyClassifier internally applies EasyPreprocessor, so no preprocessing is necessary.
- Parameters:
- n_jobsint, default=None
Number of processes to spawn for parallelizing the search.
- min_resources{‘exhaust’, ‘smallest’} or int, default=’exhaust’
The minimum amount of resource that any candidate is allowed to use for a given iteration. Equivalently, this defines the amount of resources r0 that are allocated for each candidate at the first iteration. See the documentation of HalvingGridSearchCV for more information.
- verboseinteger, default=0
Verbosity. Higher means more output.
- type_hintsdict or None
If dict, provide type information for columns. Keys are column names, values are types as provided by detect_types.
- portfoliostr, default=’baseline’
Lets you choose a portfolio. Choose ‘baseline’ for multiple classifiers with default parameters, ‘hgb’ for high-performing HistGradientBoostingClassifiers, ‘svc’ for high-performing support vector classifiers, ‘rf’ for high-performing random forest classifiers, ‘lr’ for high-performing logistic regression classifiers, ‘mixed’ for a portfolio of different high-performing classifiers.
- Attributes:
- search_HalvingGridSearchCV instance
Fitted HalvingGridSearchCV instance for inspection.
- est_sklearn estimator
Best estimator (pipeline) found during search.