Method
Feature selection
Selecting features by filters, wrappers, Boruta, RFE or mutual information.
21 cases from 18 companies apply Feature selection. 2 were checked in an online A/B test; 8 report results in production.
Typical tasks: Catch fraud and abuse, Optimize decisions, Evaluate models, Moderate content at scale and 9 more. Companies: VTB, Uber, Affirm, Instacart, Alfa-Bank, Target and 12 more. Often combined with: Gradient boosting, SHAP, Lag and window features, Logistic regression, Imbalanced learning and 76 more.







