Journal Article

Artificial Intelligence Modeling of Mood, Coping, Work Engagement and Social Factors in Predicting Mental Health Outcomes

Bhattacharya Anupam, Ghosh Sayani, Ghosh Saugata, Kundu Prasenjit

2025

Overview

Mental health conditions currently account for approximately 14 percent of the global disease burden, according to the World Health Organization, while depression will be the leading cause of disability by 2030. This developing crisis underlines the pressing need to ascertain behavioral and psychological markers associated with mental health risks. This study investigates psychosocial determinants of self-reported mental health history by analyzing a large-scale publicly available behavioral dataset comprising 292,364 individual observations. Artificial Intelligence (AL) based machine learning logistics Modeling framed in this study with self-reported mental status as response variable and four key psychosocial predictors: Mood Swings, Coping Struggles, Work Interest and Social Weakness as independent variables. Initial bivariate models indicated significant associations between Mood Swings and Coping Struggles and the likelihood of reporting mental health conditions. In the final logistic model, Mood Swings (p < 2e-16, T value: 17.61) and Work Interest (p < 2e-16, T value: 31.48) emerged as the most significant positive predictors, while Social Weakness showed a statistically significant negative association. VIF scores indicated no multicollinearity among the predictors. Further, gender-stratified modeling framed by this current study showed striking differences in predictor behavior between male and female respondents; this is especially true for social functioning, which appeared positively significant in females but negatively significant in males, and for coping mechanisms, which appeared much stronger in females. These findings point toward the important role of emotional regulation, managing of stress, vocational engagement, and interpersonal dynamics in shaping mental health outcomes. This brings into evidence the optimal threshold on the ROC curve, depicting that about 70.4% of the actual positive cases of mental health are correctly identified by the model. The logistic model slightly outperformed and revealed that AI can be used to predict mental health risks.

• Publisher: EDP Sciences
• Subject: artificial intelligence modeling / behavioral analysis / coping mechanisms / mental health / mood swings / psychosocial predictors