Process Flow

The following tasks are used for creating a Learn Model by using the Learn UI application:

Task

Operation

Creating, Saving, and Opening a Project

Create a new project and add a data file to the project.

Reviewing List of Fields

Select the key field and assign field types. The statistics and distribution of values displayed by the Learn UI can determine the appropriate field type and the importance of the field in matching.

Defining Features

The set of features provides all the information on which the model bases its decisions. Each feature is a comparison of particular field values in a record pair.

Identifying Record Pairs

This step involves creating the datasets used to train and validate the model by finding record pairs that represent many possible data combinations and assigning “True” or “False” labels to the pairs. Useful record pairs can be found automatically.

Training a Learn Model

The model is trained using a number of iterations. Statistics of the trained model are displayed. The trained model can be saved together with model scores for each record pair. Features and pairs can be modified based on the provided suggestions.

Testing a Trained Model

This provides the ability to evaluate the accuracy of a trained model with record pairs that were saved with the current or another model or project.

Saving and Reviewing Trained Models

All saved models are displayed. You can export the trained model to be loaded to a TIBCO Patterns server. It can be used to find matching records in a TIBCO® Patterns search table that has the same structure as the training data.

Figure 2: Learn UI Process Flow