Dynamic Ensemble Selection for Mental Health Prediction : A Path towards Explainable, Scalable and High-Impact AI Solutions

Mental health conditions such as anxiety, depression, PTSD, and insomnia are widespread, impacting individuals' emotional, physical, and social lives. Conventional diagnostic methods are often expensive, time-consuming, and influenced by stigma and limited accessibility, discouraging many from seeking timely help. Early identification plays a crucial role in preventing severe developments and supporting overall well-being. In order to solve such issues, we have proposed a Machine Learning (ML)-driven system that evaluates mental health status based on responses to a thoughtfully crafted survey using clinically recognized psychological indicators. This approach leverages dynamic ensemble selection techniques like KNORA-U and KNORA-E, along with SHAP (SHapley Additive exPlanations), to enhance both accuracy and interpretability. Achieving an accuracy of 81%, our model outperforms conventional methods and highlights the potential of combining explainable AI with adaptive ensemble strategies. This framework promotes early detection, improves accessibility, and offers a scalable, reliable solution for mental health assessment and intervention.