As a developer I want to have the contribution of each L2G feature in the predictions dataset because it helps understanding the feature value in the context of the importance of each feature in the model.

Background

I see this work being beneficial in 2 aspects:

• Analysis purposes: programmatically explaining how L2G has come out with a certain prediction
• Visualization purposes: we could use these contributions in the UI of the platform, as mentioned here L2G widget on credible set page #3603

Tasks

• Calculate Shapley values for each L2G prediction and add it to the final dataset