Mapping a decade of research on employee attrition prediction: a bibliometric and thematic analysis of machine learning applications (2014–2025)

This paper synthesizes and maps ten years of research on employee attrition prediction using machine learning (ML) (2014–2025). 161 peer-reviewed English-language articles and reviews were selected from a starting set of 610 records using a structured Scopus search (title, abstract, and keywords) and a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-guided screening procedure. Both network mapping and bibliometric tools were used: VOSviewer was used for co-occurrence and co-citation mapping, while Biblioshiny was used for descriptive and thematic analysis. As a result of the interdisciplinary nature of the topic, the results show a significant increase in publications after 2021, with contributions mostly appearing in management journals and technical publications like IEEE Access and Expert Systems with Applications. Four groupings emerged from keyword and topic analyses: (1) predictive methods and machine learning approaches; (2) human resource notions, including employee retention and turnover intention; (3) explainable AI and fairness (e.g. SHapley Additive exPlanations (SHAP), Local Interpretable Model-agnostic Explanations (LIME); and (4) educational, information technology, and healthcare sector-specific applications. Co-citation networks demonstrate a dual intellectual foundation that combines fundamental ML literature with traditional human resource management turnover ideas, which are connected by new interdisciplinary applications. The results emphasize the necessity of cross-sectoral cooperation, broader implementation of interpretable ML in human resource practice, and more robust theory-driven model building. The paper offers a thorough road map for developing ethical, theoretically grounded attrition prediction research.