Interpretable Deep Learning for Psychosocial Assessment of Women's Mental Health and Empowerment

Abstract Women’s mental health is an important public health concern that significantly impacts their overall well-being and quality of life. Among several mental health challenges, depression is one of the most common conditions affecting women, while maternal depression presents additional psychological challenges during pregnancy and the postpartum period. These conditions are influenced by multiple social, economic, psychological, and empowerment-related factors, making their accurate assessment challenging. Machine learning (ML) and deep learning (DL) approaches are offers effective means of identifying hidden patterns and predicting women’s mental health outcomes. Moreover, existing methodologies have limited capability to explain how empowerment and psychosocial factors influence women’s mental health, highlighting the need for an explainable framework. To overcome these challenges, this research work develops an Explainable Women's Mental Health and Empowerment Assessment (X-WMHEA) model using psychosocial factors for effective analysis. The proposed model utilizes HSIC-LASSO to identify informative and non-redundant psychosocial factors by recognizing their dependencies with the target outcomes. Moreover, the TabM model learns complex relationships among the selected psychosocial factors and classify women’s mental health conditions. In addition, the Adan optimizer is utilized to refine the model parameters and enhance the convergence of the learning process. To enhance model transparency, the SHAP is integrated to quantify the contribution of individual psychosocial factors and provide meaningful insights into factors associated with women's mental health and empowerment. The performance of the proposed X-WMHEA model demonstrates superior performance, achieving accuracy of 95.53% and 96.75% on the Women Depression Dataset and Maternal Depression Dataset. Additionally, an ablation study is conducted to examine the contribution of each component in the proposed framework. Thus, the proposed model provides an effective and interpretable approach for analyzing psychosocial factors associated with women’s mental health and empowerment, supporting reliable assessment and informed decision-making.