Predicting Academic Rank Improvement of Higher Education Lecturers Using Machine Learning Model

This study aims a spirited predictive model to determine lecturers who have conditions for the apologize of academic rank based on key performance indicators like teaching evaluations, research productivity, and professional contributions. Data modeled included 1,229 lecturers and compared four machine learning algorithms as Random Forest, Support Vector Machine, Logistic Regression, and k-Nearest Neighbor. SMOTE oversampling (Synthetic Minority Oversampling Technique) was performed due to considerable class imbalance that led to an effective balanced dataset. Moreover, features that were not correlated were removed in order to improve the interpretability and performance of the proposed model. The results showed that the combination of Random Forest and SMOTE resulted in maximum performance across all cases, with an F1-score of 0.989 and accuracy of 98.9%. According to the SHAP analysis, the five most critical variables in predicting faculty performance, in order of importance, were the H-Index, publications, teaching evaluations, citations, and years of experience. Based on machine learning and explainable AI techniques (SHAP and LIME), this study was conducted to showcase using fairness, transparency and accuracy to predict the promotion for academic rank which can illustrate practical usages for higher educational institutions.