Machine Learning Applications in Employee Turnover and Performance Prediction: A Systematic Literature Review

Organizations are increasingly adopting data-driven methods in human capital management, particularly for predicting employee performance and turnover, due to the growing availability of workforce data. Although machine learning has shown promise in this field, current research often examines performance and turnover prediction independently. However, employee performance plays a critical role in predicting turnover, as these two outcomes are interconnected—performance frequently influences individual decisions to remain with or leave a company. This study aims to identify frequently used models, input features, and evaluation metrics in machine learning, as well as to evaluate models that perform well on both prediction tasks. In accordance with PRISMA principles, a systematic literature review (SLR) of peer-reviewed research published between 2020 and 2025 was conducted. After applying predetermined inclusion and exclusion criteria, 23 papers were selected from academic databases (Scopus and Google Scholar). The review results demonstrate that tree-based ensemble models and neural network-based methodologies frequently outperformed other machine learning methods in predicting both performance and turnover. Models such as Random Forest, Gradient Boosting, XGBoost, and deep learning architectures (DNN, RNN-LSTM) delivered exceptional prediction accuracy, while hybrid models further enhanced outcome explainability, reliability, and practical application in HR analytics. These findings offer critical managerial implications by guiding HR practitioners in selecting appropriate predictive models for workforce planning, retention strategies, and performance management systems. They also highlight key research gaps, including dataset bias, ethical considerations in algorithmic decision-making, and the need for longitudinal validation studies across diverse organizational contexts.