Modelling the Dependence of Employee Burnout on Their Expectations Satisfaction by Machine Learning Methods

As part of the current task of predicting employee burnout, machine learning models are being developed to predict burnout based on data regarding the fulfillment of employee expectations from the corporate well-being program. The source data consists of survey results from employees of large companies. To predict the degree of burnout, a classification model based on fuzzy levels of burnout is built. The correspondence matrix between the fuzzy ranges of the integral indicator of expectation fulfillment and the fuzzy levels of burnout is optimized using the criterion of entropy minimization. The task of binary classification for predicting the presence of burnout is also addressed. For this purpose, a set of rules is formed that provides an explanation for the machine learning model. Each rule uses a couple of features. The machine learning model includes 10 decision rules and achieves an accuracy of 80% for burnout prediction. Based on the constructed model, it is concluded that the nature of burnout differs depending on the implementation of corporate well-being activities for different clusters of employees based on expectation level. At the same time, the assignment of an employee to a particular cluster is related to his value priorities. Thus, the study allows for the identification of hidden factors that are determined by the values of employees and affect their burnout.