Abstract:  In the era of digital economy, organizations need to invest a lot of resources in developing employees’ digital skills to maintain a competitive advantage; however, organizations face the risk of employee turnover when they succeed in developing highly-skilled employees. This paradox between talent cultivation and talent turnover has not been properly resolved for a long time. The study aims to provide new perspectives to solve this problem by predicting whether employees will leave and constructing a portrait of retained employees through machine learning techniques with the help of XGBoost model and K-means cluster analysis. The study is based on the voluntary redundancy model and is conducted by using an open source employee turnover dataset containing 19,158 sample points on the Kaggle data analytics platform. The study shows: (1) There are four types of retained employee portraits, including entry-level, women with arts-business-humanities background, senior-stability, and opportunity-chasing; (2) Employee screening plans and training strategies are proposed based on the four types. The findings reveal the potential of machine learning techniques in building accurate employee retention prediction models, which are of theoretical reference and practical guidance for organizations to screen employees with training value before implementing digital skill training and to make scientific employee retention decisions.
Key words:  machine learning; organizational talent development paradox; digital skill training; employee turnover portrait   北京邮电大学学报（社会科学版）2024年第4期第26卷第4期2024年8月北京邮电大学学报（社会科学版）Journal of Beijing University of Posts and Telecommunications（Social Sciences Edition）Vol.26，No.4Aug.2024

Key words: machine learning, organizational talent development paradox, digital skill training, employee turnover portrait