Machine Learning for Human Resource Management in the Public Sector: Early-Career Turnover Intention Prediction Model

Abstract

The study aims to develop an optimal turnover intention prediction model through training, including explanatory variables that have causal relationships with early-career turnover intention in the public sector. This study employs job-related factors, organization system factors, and socio-psychological factors, using the Public Employee Perception Survey data. According to the result, the random forest prediction model provides a better fit to predict early-career turnover intention and most variables of the importance index of turnover intention are P-O fit, P-J fit, satisfaction with the welfare benefits system, and HR, job satisfaction, autonomy, education, and prospects. This study provides policy implications and future direction to decrease early-career turnover intention in the public sector in South Korea.