Classical Machine-Learning Classifiers to Predict Employee Turnover

Abstract

In the context of the increasing migration of employees from one company to another and given the career changes that many people desire in order to avoid monotony, the dynamics of the workforce puts great pressure on the stability of a company. HR departments can establish more effective mechanisms for preventing employee turnover and better recruitment strategies when given a reliable Machine Learning tool. Facing a classification problem (employees are either tagged as “left” or “not left”), this investigation attempts to conduct experiments with traditional methods such as Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine to predict employee turnover. In our analysis Support Vector Machine showed promising results in predicting employee turnover. We conclude that we obtained sufficient results in order to trust a Machine Learning classifier to label correctly the employees that left the company and to advise practitioners to integrate such tools into the everyday activity of the HR department.