Prediction of Mental Burnout Using Machine Learning

Burnout has become a major issue in the workplace and in academia, along with the potential risk mental health repercussions and performance consequences. Those suffering from burnout will typically experience one of three symptoms: reduced personal accomplishment, depersonalization and emotional exhaustion. These three psychosocial factors overall impact productivity, psychological health and business effectiveness. Traditional methods for assessing levels of burnout typically utilize retrospective, small-n methodologies such as surveys and interviews, which do not necessarily measure real-time indicators. The current research employs a predictive model using machine learning (ML) that can reveal burnout patterns in workers based on their work habits and lifestyle behaviours.The dataset was obtained from Kaggle and included nearly 3,000 survey responses from employees. To overcome the imbalance of class, the study used the Synthetic Minority Over-Sampling Technique method prior to training a Logistic Regression model. The model received an accuracy score of 99.6% and strong predictive ability in conjunction with cross-validation. The study offers possibilities for discovering burnout, in a proactive manner, using ‘ML-related’ systems. The article explains data pre-processing, model validity and discusses the potential of ML to help elucidate mental health-related systems within the workplace context.