Package com.tibco.patterns.learn.api.hint

Generation of all types of model training suggestions, filtering of data table for a specific subset in a suggestion, analysis of datasets by subset and label.

The model training suggestions (hints) are generated after a trained model has been saved. All suggestions implement the Hint interface.

There are two main types of hints: suggestions that ask to add more pairs to specific subsets (descendants of HintAddPairs) and suggestions that ask to review specific existing pairs (descendants of HintReviewPairs).

Each hint type has a static create() method that examines the state of the model and creates an instance of the specific Hint subclass if this suggestion is applicable. A null value is returned if the suggestion is not applicable to the current state of the model.

Hints to add more pairs have a family of specific RLink subsets that the pairs should be added to. The SubsetFamily class provides functionality to filter data table records that correspond to the specific RLink subset, thus assisting in selecting suitable record pairs for each suggested subset.

This package also contains DatasetStats and SubsetLabelPairMap classes that calculate statistics of a given dataset and help identify whether a new record pair is likely to be useful for training the model.