Drug usage classification based on personality and demographic features using a combination of sampling and machine learning algorithms.

Drug use stems from biopsychosocial factors. This study classified 18 drug types using personality and demographics. After preprocessing, three sampling techniques Random Oversampling, Synthetic Minority Over-sampling Technique using Euclidean Norm (SMOTEN), Synthetic Minority Over-sampling Technique using Euclidean Norm - Edited Nearest Neighbors (SMOTEENN) and seven machine learning (ML) models Random Forest (RF), Extreme Gradient Boosting (XGBoost), Decision tree (DT), Extra Tree, Support Vector Classification (SVC), Linear SVC and Logistic Regression (LR) were applied to build a robust, accurate prediction model for drug classification. Random Over Sampler and Extra Trees improved F1 scores in unbalanced data, as shown in a case study.