How to Use Learn Models to Evaluate Feature Vectors

You can use the Learn model for evaluating any feature vectors computed by one or more external applications. You can use the Learn model to detect hidden patterns in some particular problem domain, not tied exclusively to text or record matching. For example, fraud detection or spam detection might depend on various metrics depending on a user’s pattern of interaction with a given system. The features in the feature vectors used for evaluation must be the same as the features used to train the model.

Two types of functions are required for feature vector evaluation:

•

Loading and managing in-memory Learn models.

•

Evaluating feature vectors using a selected in-memory Learn model. This is done using a function that obtains the score computed using the Learn model for the given feature vector.