9. Create Feature List
You'll generate a new feature list by prioritizing features with higher predictive scores, then seek additional signals not captured in the initial list.
Step 1: Create a New Feature List¶
Navigate to the Feature catalog from the Experiment section of the menu.
Sort the Features by their predictive scores.
Add the top 10 features to the Feature List Builder by clicking on the 
icon for each top 10 feature.
Click on 
in the Feature List Builder to further explore and refine the selection.
This should take you to a window with the list details.
Click on any feature to review its details.
Review the semantic relevance by Generative AI and edit it if desired.
Save Feature List as "Top10 Most Predictive for Activity Next Week" by clicking on 
at the upper right of the window.
Our new feature list is now part of the Feature List Catalog.
Step 2: Evaluate Feature List¶
Click on the feature list we just created to review its details.
Go to the 'Themes' tab to assess any missing themes in the list. By selecting the top 10 features, we left a lot of signal on the table. Although the feature list contains the top 10 most predictive features, a feature list with more diverse signal is likely to lead to better models.
Step 3: Create a Diverse Feature List¶
Close window and click on Features 
to return to the Feature Catalog.
Sort the Features by their predictive scores.
Add the feature with the highest predictive score to the Feature List Builder by clicking on the icon of this feature.
The Feature List Builder now incorporates one feature.
Review the Builder's suggestions by clicking on themes in the 'Suggestions' section located at the bottom of the Feature List Builder.
Click on 
for the Customer/InvoiceItems/Frequency theme to explore features associated with this theme.
Add the most predictive feature to the Builder by clicking on the icon of this feature.
We will repeat this for each missing theme:
- Customer/GroceryInvoice/Stability
- Customer/GroceryInvoice/Stats
- Customer/GroceryInvoice/Diversity
- Customer/GroceryInvoice x InvoiceItems/Stability
- Customer/InvoiceItems/Stats
- Customer/GroceryInvoice/Recency
- Customer/GroceryInvoice/Timing
- Customer/InvoiceItems/Bucketing
Only for Customer/InvoiceItems/Diversity we will add a specific feature "CUSTOMER_Entropy_of_items_TotalCosts_by_product_ProductGroup_4w" to the Feature List Builder.
We have now 11 features in the Feature List Builder.
Click on and save Feature List as "List with diverse signals".
The enhanced Feature List is now accessible through the Feature List Catalog.
You can iterate by adding and removing features. You can also explore more features from Feature Ideation to enhance your feature list.