Machine Learning Approaches for Mental Health Prediction

Mental health issues within workplace environments constitute a significant global concern. According to the World Health Organization (2024), anxiety and depression together contribute to an estimated loss of over one trillion U.S. dollars in productivity each year. Although awareness of this problem has grown across organizations, most existing initiatives remain reactive, intervening only after symptoms appear rather than proactively identifying individuals at risk. The present research employs a predictive and data-centric methodology to anticipate employee vulnerability to mental health challenges, drawing on the Mental Health in Tech Survey dataset hosted on Kaggle. Guided by the CRISP-DM process model and supported by the job demands resources (JD–R) theoretical framework, the study incorporated comprehensive data preprocessing, feature construction, and machine learning techniques. To evaluate model effectiveness, several performance metrics were applied, such as accuracy, F1-score, precision, recall, and ROC-AUC, along with a feature importance analysis that connected statistical patterns to conceptual insights. Among the tested algorithms, the Random Forest model demonstrated the best results, achieving 87.2% accuracy and an ROC-AUCof 0.93. Key predictors identified included organizational support and familial history of mental health conditions. These findings highlight the value of combining psychological theory with computational modelling to generate actionable intelligence. Such an approach can inform early detection strategies, help destigmatize mental health concerns, and promote the development of more supportive and resilient workplace cultures.