AI Model Predicts Mental Health Risks for Female Sex Workers

This research introduces a novel hybrid predictive model designed to identify mental health risks, specifically depression, within marginalized groups like female sex workers. The model integrates an ensemble feature selection strategy, utilizing ANOVA and mutual information, with a Harris Hawks optimization-tuned logistic regression. This approach represents a new application of swarm intelligence in predicting mental health outcomes for vulnerable populations. The study applied this model to a dataset of 3,005 female sex workers, achieving high accuracy (95.78%), F1 score (95.77%), and AUC (0.96). Crucially, the model incorporates explainable AI (XAI) methods, allowing for a deeper understanding of the factors contributing to mental health issues. It pinpointed post-traumatic stress, violence from clients, and specific occupational factors as significant drivers of depression. This work aims to bridge the gap between traditional and machine learning methods, offering an XAI tool that can facilitate early intervention, targeted psychosocial care, and evidence-based health planning for vulnerable individuals.