Identifying Key Predictors of Students’ Competency Achievement Using Machine Learning Models: A Bioengineering Case Study

Competency-based education (CBE) in higher education demands interpretable and scalable tools to monitor student progress. Current studies on CBE have used small samples in short evaluation periods or have not used machine learning or explainability of the results. This study introduces a robust analytical pipeline that integrates correlation analysis, Factor Analysis of Mixed Data, and explainable machine learning to predict competency achievement in bioengineering programs. Using over 300,000 evaluations from a private Mexican university, Random Forest model achieved outstanding predictive performance in a Stratified 10-fold Cross-validation experiment (AUC = 0.9604–0.9653), outperforming deep neural networks for One-Class Classification in highly imbalanced data. Model interpretability using SHAP highlighted academic and course-related variables, rather than demographic factors, as the strongest predictors, reinforcing the fairness of the evaluation process. This work advances the operationalization of explainable AI in CBE, contributing to the emerging vision of Data-Based Education by providing actionable insights for curriculum design, academic advising, and institutional policy.