Predicting Employee Burnout Risk Using Machine Learning and Workplace Well-Being Indicators

Abstract

Employee burnout has become a critical concern in human resource management, particularly in flexible and work-from-home environments where workload intensity, digital exposure, and work-life boundaries are increasingly difficult to monitor. This study aims to develop a machine learning-based model for predicting employee burnout risk using daily work-from-home behavioral patterns and workplace well-being indicators. The dataset consisted of 1,800 daily records collected from 180 users, including work hours, screen time, meeting frequency, breaks, after-hours work, sleep duration, task completion rate, burnout score, and burnout risk category. To prevent target leakage, user identity and burnout score were excluded from the predictor set. The original burnout risk label was transformed into a binary classification target by grouping Medium and High categories into an At-Risk class, while Low was retained as Low Risk. Several supervised machine learning algorithms were evaluated using 10-fold stratified cross-validation, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosted Trees, Support Vector Machine, Naïve Bayes, and k-Nearest Neighbors. The results showed that Random Forest achieved the best overall performance, with 97.11% accuracy, 94.37% macro-F1, 94.24% balanced accuracy, and 90.11% recall for the At-Risk class. Feature importance analysis indicated that task completion rate was the most influential predictor, followed by work hours, sleep hours, and screen time. These findings demonstrate that machine learning can support HR analytics-based early warning systems for employee burnout risk detection in remote work settings.