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

Researchers developed an explainable AI framework combining GAN-based oversampling, Dragonfly Algorithm optimization, and XGBoost to predict mental health outcomes in drug-affected populations, achieving 94.17% accuracy. The model addresses class imbalance and interpretability challenges in clinical settings, identifying behavioral factors like sleep quality and emotional regulation as key predictive indicators.

This research tackles a critical gap in healthcare AI by focusing on mental health prediction within vulnerable drug-affected populations, where traditional diagnostic methods often fail due to limited access and resource constraints. The study's significance lies not in novel algorithmic breakthroughs but in demonstrating how established techniques can be systematically combined to solve a real clinical problem. The integration of GAN-based oversampling addresses the persistent challenge of imbalanced medical datasets, while the Dragonfly Algorithm provides automated hyperparameter tuning that reduces manual optimization overhead.

The framework's reliance on explainable AI through SHAP analysis reflects growing healthcare industry demands for model transparency and clinical trust. Mental health prediction in drug users remains understudied in AI literature, partly due to stigma and data collection difficulties. This work establishes a methodological template applicable beyond this population to other underserved patient groups.

For healthcare technology developers and AI researchers, this demonstrates market viability for specialized mental health prediction tools. The 94.17% accuracy suggests clinical-grade performance potential, though real-world deployment requires validation across diverse populations and healthcare systems. Investment in interpretable AI for healthcare continues accelerating as regulatory bodies demand transparency in clinical decision support systems.

The research indicates that lifestyle and behavioral factors substantially outweigh demographics in predicting mental health outcomes—a finding with implications for intervention design. Future development should focus on prospective validation in clinical environments and integration with existing electronic health records systems to determine practical effectiveness.

• →Hybrid ML framework combining GAN oversampling and Dragonfly Algorithm-optimized XGBoost achieved 94.17% accuracy in mental health prediction for drug-affected populations.
• →SHAP-based explainable AI enhances clinical trust by providing interpretable, instance-level predictions suitable for healthcare decision-making.
• →Sleep quality, physical health, and emotional regulation emerged as primary mental health predictors, while demographic factors showed minimal impact.
• →The framework effectively handles high-dimensional categorical data and addresses class imbalance—persistent challenges in medical AI.
• →Study identifies significant gap in AI-based mental health research for drug users and provides methodology applicable to other underserved populations.