Managing training data for Watson OpenScale
You must connect Watson OpenScale to your training data to configure model evaluations and explainability.
Watson OpenScale uses training data to calculate the metrics for your model evaluations and explainability methods. To configure model evaluations and explainability methods in Watson OpenScale, you must prepare and store your training data.
Preparing training data
The format of your training data can determine the results of your model evaluations. To enable model evaluations, you must prepare to store your training data in a format that Watson OpenScale can process. Your training data must contain labeled feature columns and a prediction column as shown in the following example:
Watson OpenScale uses the training data that you provide to create a training data schema to ensure that your training data corresponds with the format that it understands. The schema specifies the feature columns that you provide in your training data and the type of data that the columns contain. The following example shows a training data schema for the German Credit Risk dataset:
Storing training data
You can store your training data in a supported Db2 database and connect the data to Watson OpenScale. For more information, see Connectors.
You can also store your data in Cloud Object Storage.
If you want to keep the details of your training data location private, you can also connect Watson OpenScale to your training data by running a custom notebook to upload the configuration file that it generates.
Next steps
You must connect your training data to Watson OpenScale so that it understands how to process your model.
Learn more
Parent topic: Managing data for model evaluations in Watson OpenScale