Mental Health In Tech Survey

Abstract

The widening gap between mental health prevalence and effective workplace support in the technology sector necessitates proactive risk detection mechanisms. Addressing this need, we engineered a machine learning framework to predict work-interfering mental health risks by leveraging a dataset of over 1,200 tech professionals. We benchmarked four classifiers, specifically Logistic Regression, Decision Tree, Random Forest, and Gradient Boosting, and treated the challenge as a binary classification task where sensitivity is paramount. The Random Forest model emerged as the optimal solution by securing a Recall of 0.878 and an F1-score of 0.803, thereby ensuring robust identification of at-risk employees with minimal oversight. Crucially, our fairness audit confirmed that the model maintains performance parity across gender demographics. Beyond prediction, feature analysis revealed that organizational factors like supervisor support and the clarity of leave policies act as primary determinants alongside individual treatment history. These findings advocate for a shift from reactive measures to systemic interventions and offer organizations a scalable detection tool coupled with evidence-based levers to actively improve workforce well-being.