Method
Gradient boosting
Boosted decision trees (CatBoost, LightGBM, XGBoost).
145 cases from 74 companies apply Gradient boosting. 42 were checked in an online A/B test; 46 report results in production.
Typical tasks: Optimize decisions, Rank search results, Catch fraud and abuse, Forecast demand and load and 22 more. Companies: Yandex, Avito, DoorDash, VTB, Yandex Lavka, OLX and 68 more. Often combined with: A/B test, SHAP, Threshold selection, Rule-based heuristics, Embedding similarity matching and 223 more.










