Approaches :

1- Train different models (LGBM, CatBoost,XGBOOST) the best one was CatBoost

2-Parameters Tuning but It didn't work because there is no correlation between local validation and the LB

3- Find the best Cross-validation strategy the best one is Kfold with 10 folds

4- training with different seeds and average the predictions

https://github.com/JedidiMohamed/Second-place-solution-Womxn-in-Big_Data_South_Africa