Swarm AI model predicts depression in female sex workers with 96% AUC

A new paper presented at the 2026 IEEE Symposium on Computers & Informatics introduces a novel hybrid model for explainable mental health risk prediction in female sex workers (FSWs). The model combines ensemble feature selection using ANOVA and mutual information with a logistic regression classifier tuned via Harris Hawks Optimization (HHO), a swarm intelligence algorithm. Tested on a dataset of 3,005 FSWs, it achieves 95.78% accuracy, 95.77% F1 score, and an AUC of 0.96, outperforming traditional classifiers. Explainable AI (XAI) techniques reveal that post-traumatic stress, client-related violence, and occupational hardship are the most significant risk factors for depression in this marginalized group.

The work directly addresses the challenge of applying machine learning to high-dimensional, complex risk patterns in vulnerable populations. By integrating swarm intelligence (HHO) with classical feature selection, the model maintains interpretability while boosting predictive performance. The authors emphasize that this XAI tool can bridge the gap between conventional mental health assessment and data-driven approaches, enabling early detection and evidence-based psychosocial care. The model is designed for deployment in low-resource settings, helping clinicians and policymakers target interventions for FSWs experiencing trauma, violence, and economic instability. The paper appears in IEEE conference proceedings and is available on arXiv.