Use of unique RandIDs Limited to 2K datapoints Fixed cigperday Transformation for skewness of Cigperday skewness Scaling of Age and Cigperday Exclusion of outliers Encoding categorical features. RFE better features and metrics Features by RF, byLasso, by Correlation clight increase of measures with RFE, increase of specificity: ## Accuracy Precision Recall Specificity F1_Score AUC ## Accuracy 0.7983871 0.4888889 0.2222222 0.9420655 0.3055556 0.7611633

Unique RandIDs

Unique IDS

Code for decision tree (xgboost), random forest (xgboost), and logistic regression. Code also included for 2 ensemble stacks, decision tree _ logistic regression, and random forest + logistic regression.