Build Feature Stores for Predictions

The Feature Stores section of the Workbench is where you store, manage, and retrieve precomputed features, ensuring consistency and reusability across various machine learning models and applications.

Using the collection you have just exported, you will need to create a feature store to be used in your recommender.

Manage Feature Stores

In the Recommenders section of the Workbench, you will find Feature Stores. Here, you will be able to add, view, and delete all the feature stores.

Add a Feature Store

To create a feature store from your exported data set, click + Feature Store. The input fields will be where you set the configuration settings for your final Feature Store. You will need to provide a unique name in Feature Store ID and a good Description. The feature store can also be allocated to an existing project to help ease your project creation journey.

Check Data

Specify the characteristics of the file you will be building the feature store on, and click Check Data. Clicking Check Data will create a table of all of the columns in your data set. You can change the automatically allocated type of each column on the right. You can also add descriptions to those columns if you wish. Once you have correctly typed and described the details, the Destination Frame will be pre-populated with your feature store name .hex file. If you want your feature store to have a different name, you can change it here. This .hex file will be used in the next step when configuring your predictions.

Parse Data

To trigger the creation of your feature store with all these details, click Parse Data.