After you identify the match model for training, configure the match fields that the machine learning (ML) model requires for the training process.
You can review the match fields before you start training the ML model. The ML model uses the match fields that you configure for the declarative rules, candidate selection criteria, or both. Apart from the existing match fields, you can add additional match fields to train the ML model.
Vew the following details for the match fields that are part of the training:
•Field Name. Name of the match field.
•Field Path. Complete path of the match field. If the field does not have a parent or path, the Field Path column does not display any value.
•Internal ID. Unique internal identifier for the match field.
•Used in. Information about where the match field is used. The supported values are declarative rules, candidate selection criteria, and training.
If you add a match field, the Used in column displays Training as the default value.
You can delete the match fields that you add. You cannot delete a match field that is part of declarative rules or candidate selection criteria.
Match fields for a pure ML approach
If you adopt a pure ML approach, your match model must not contain any declarative rules. The ML model uses the match fields that the candidate section process uses. The match fields that you configure as part of the candidate selection criteria impact the training and determine the accuracy of the ML model.
Match fields for a hybrid ML approach
If you adopt a hybrid ML approach based on a set of declarative match rules and an ML model, your match model must contain declarative match rules and candidate selection criteria. The hybrid model uses the match fields that the declarative match rules and candidate section process use. The match fields that you configure impact the training and determine the accuracy of the ML model.