Multimodal Analysis and Predictive Modeling for Mental Health Detection through Lifestyle and Behavioral Data using Machine Learning

Abstract

Mental health disorders—including depression, schizophrenia, and anxiety—have emerged as critical crises worldwide with rising prevalence, yet remain underdiagnosed due to limited early detection methods outside clinical settings and restricted access to psychological services. In response, my study proposes a data-driven approach for detecting and analyzing mental health risks by integrating lifestyle factors (such as dietary habits, sleep, and physical activity), behavioral indicators (stress, anxiety, social support), and demographic variables(heatmaps, boxplots, and pair plots) to uncover key relationships through a comprehensive machine learning-based framework. Utilizing a comprehensive set of visualizations—including heatmaps and histograms—and a linear regression model, the framework accurately predicts depression scores, demonstrating a strong correlation between actual and predicted values. Significant trends included strong positive correlations between stress, anxiety, and depression scores, and inverse relationships between mental health risk and variables like sleep and social support. Gender-based comparisons further revealed that women reported higher depression scores and suicidal ideation than men. To predict mental health risks, various machine learning models were evaluated, with a Random Forest classifier emerging as the most effective, achieving strong performance in classifying individuals into low, moderate, and high-risk categories. Feature importance analysis showed that stress level, anxiety score, depression score, and sleep hours were the most influential predictors of mental health outcomes. These insights support the potential of machine learning to assist in early detection and personalized intervention planning. My findings lay a foundation for the development of scalable, interpretable, and accessible tools for mental health screening and underscore the value of integrating machine learning technology and behavioral datasets into public health strategies thus offering a scalable approach for early detection.