Tabular ML (bookmark)

Supporting table-level modelling after a habitat FeatureTable. This is not the product core and is not a wall of classifiers on API Reference.

User guide: Table ML domain API. Composer: TablePipeline (documented on habit.pipeline: compose subject / table pipelines). Construct components with each domain registry’s create / constructor_signature (do not invent sklearn-shaped aliases). Image-side feature scaling is habit.feature_preprocessing: scale voxel-feature matrices.

Classes

habit.table_preprocessing.TablePreprocessor	Learn and apply a feature-table transformation.
habit.table_preprocessing.TablePreprocessorRegistry	Name-to-implementation registry for feature-table preprocessors.
habit.feature_selection.FeatureSelector	Learn and apply a feature-column subset.
habit.feature_selection.FeatureSelectorRegistry	Name-to-implementation registry for feature-table selectors.
habit.classification.Classifier	Outcome model over feature tables.
habit.classification.ClassifierRegistry	Name-to-implementation registry for outcome classifiers.
habit.evaluation.Metric	Evaluation metric with explicit input requirements.
habit.evaluation.MetricRegistry	Name-to-implementation registry for evaluation metrics.
habit.recipes.ModelResult	Outcome of train_model(), entirely in memory.
habit.recipes.CVResult	Outcome of cross_validate(), entirely in memory.
habit.recipes.PredictionResult	Outcome of predict_model(), entirely in memory.
habit.recipes.SearchResult	Outcome of search_hyperparameters(), entirely in memory.

Functions

Recipe helpers exported from habit.recipes:

habit.recipes.train_model	Fit one pipeline on a table and score it, with an optional hold-out split.
habit.recipes.cross_validate	Estimate generalisation with stratified K-fold cross-validation.
habit.recipes.predict_model	Apply one FITTED pipeline to new rows (the inference recipe).
habit.recipes.search_hyperparameters	Tune hyperparameters by K-fold search and write the winners into the spec.
habit.recipes.compare_models	Compare multiple trained models from a validated comparison config.
habit.recipes.pairwise_delong_test	Compare two models' ROC AUCs on the same subjects (paired DeLong test).
habit.plugins.create_ml_model	Build one ML model instance for config validation (habit check-config).

Evaluation statistics

habit.evaluation.auc_confidence_interval	Compute the ROC AUC and its DeLong confidence interval.
habit.evaluation.calibration_tests	Run both calibration tests HABIT reports for a binary model.
habit.evaluation.delong_test	Compare two models' ROC AUCs on the same subjects (paired DeLong test).
habit.evaluation.icc_analysis	Compute per-feature test-retest ICCs across measurement sessions.
habit.evaluation.repeat_measurement_matrix	Build one feature's (subjects x sessions) repeat-measurement matrix.
habit.evaluation.AucConfidenceInterval	ROC AUC with its DeLong confidence interval.
habit.evaluation.CalibrationResult	The two calibration test outcomes reported together for a model.
habit.evaluation.DelongResult	Outcome of a paired DeLong AUC comparison.