Employee Turnover Prediction with Supervised Machine Learning

Abstract

Employee turnover is a pervasive issue faced by organizations across various industries (Korff et al., 2015). When employees decide to leave a company, it not only disrupts daily operations but also imposes costs associated with recruiting and training new talent (Cascio & Boudreau, 2011; Matthew & Kung, 2007). As a result, accurately predicting and understanding employee turnover has become a critical objective for many businesses. The advent of machine learning has opened up new possibilities for predicting and analyzing employee turnover. By leveraging vast amounts of data, organizations can develop sophisticated models that effectively forecast the likelihood of an employee leaving the company. These predictive models enable organizations to proactively address turnover risks, devise targeted retention and succession strategies, and create a more stable and successful work environment (Chanodkar et al., 2019; Perryer et al., 2010). This thesis primarily focuses on tackling the challenge of employee turnover prediction using supervised machine learning models. By utilizing four widely recognized supervised learning models, namely Decision Tree (DT), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and Extreme Gradient Boosting (XGBoost), we conduct a comparative analysis to determine the model that demonstrates the highest predictive power for employee turnover. Through our analysis, we aim to contribute to the field of employee turnover prediction by advancing the understanding of the factors that drive turnover and developing effective prediction models.