My quick take on matching algorithms using RF, XGBoost on an imbalanced dataset.
The data set is about matching players and subjects. The target is the feature "like", which happens when a player likes a certain subject.
I used random forest and xgboost to deal with the imbalanced dataset. The models in the notebook are not yet fully optimized. Regardless, they should suffice for this exercise.
Next steps would be to try stacking, to see if there is a significant boost in AUC and accuracy.