Optimizing employee promotion predictions using machine learning

Abstract

Employee promotions are essential for both organizational growth and individual career advancement, yet they often face challenges such as data imbalances and the lack of effective predictive frameworks. This study addresses these issues by applying advanced machine learning models to improve decision-making in human resource management. Using a dataset comprising 54 808 employee records, the study evaluates eight models, including Random Forest, Logistic Regression, SVM, AdaBoost, XGBoost, Gradient Boosting, Decision Tree, and Artificial Neural Networks (ANN). RF and XGBoost emerged as the most effective, with Random Forest achieving an accuracy of 96.21% and XGBoost closely following at 95.96%. Both models demonstrated strong AUC-ROC scores, highlighting their ability to handle complex data patterns. Key features influencing promotion outcomes, such as “Previous year rating” and “Average training score”, were identified as critical variables. Advanced balancing techniques such as SMOTE further improved the detection of underrepresented promoted employees, contributing to fairer evaluations. The study’s comprehensive framework, which includes detailed feature analysis and mathematical explanations, provides a practical guide for HR systems seeking to optimize promotion processes. Future research could explore hybrid deep learning models like LSTM and CNNs to enhance scalability and predictive power. Additionally, incorporating factors like employee demographics and ethical considerations would foster fairness and transparency in promotion practices, broadening the application of machine learning in HR management.