# First place solution for RF Malware Detector

Thanks
[@hayabaqais](https://www.kaggle.com/hayabaqais)
for organizing this event!
My best result is single Random Forest model with heavy ensembling of oof predictions.
Feature engineering
No cleaning required, as there are no missing data
Tried to drop column 'id', but this doesn't work, as it is highly correlated to our target
Tried to create more features (summation between two features), but this doesn't work too
Model selection
Random Forest has the best accuracy among list of common classifiers, such as LightGBM, CatBoost, KNeighbor, XGBoost
Fine tuning parameter with GridSearchCV
Key improvement
Tried out different number of KFold cross validation to generalize the model, selected FOLDS=10
Use out-of-fold prediction on test dataset
Tried out different seeds for StratifiedKFold and model
I have published my notebook
[here](https://www.kaggle.com/code/alvinleenh/malware-detector-random-forest-baseline)
. Please leave comment for further improvements or upvote 👍 if my baseline solution helps