Explainable AI for Mental Health Prediction in Drug-Affected Populations with Dragonfly Algorithm and GAN Oversampling

The study presents an explainable AI framework for multiclass mental health prediction in drug-affected populations, utilizing a hybrid approach that combines PCA-Information Gain for feature selection, GAN-based oversampling, and a Dragonfly Algorithm-optimized XGBoost model. The model achieves a predictive accuracy of 94.17% and a weighted F1-score of 93.80%, significantly outperforming traditional methods. This framework enhances interpretability through SHAP-based explainable AI, making it suitable for clinical applications and potentially improving early intervention strategies for mental health in drug users.