Abstract

In our tech-driven world, where depression affects over 5% of the global population, early detection and intervention are vital. With 1 from both 3 and 5 women and men are expected to be experiencing major depression, machine learning presents promising avenues for prediction and intervention. Depression's complexity challenges accurately diagnosing and accurate treatment, which prompt the development of various machine learning techniques to improve the management. So, this study employs five machine learning algorithms to predict depression, anxiety and stress severity using dataset from the DASS 21 questionnaire across diverse demographics. Despite class imbalances, the forest classifier considered as the most accurate and trusted model, bolstered by the incorporation of the F1 measure. Additionally, the algorithms exhibit sensitivity to negative outcomes, underscoring their potential in enhancing mental health care. This research contributes to advancing predictive models for psychological disorders, facilitating timely interventions and ultimately improving mental health outcomes in our modern society.