Higher education institutions generate large volumes of academic data; however, these data are often underutilized for proactive decision-making and early identification of at-risk students. This study proposes an intelligent academic records framework that integrates ensemble machine learning within a microservice-based architecture to address this limitation. The methodology combines Random Forest and XGBoost models, leveraging both robustness and predictive capability, while transforming academic records into structured feature vectors using performance-based and progression-based indicators such as GPA, completion ratio, failure rate, and GPA trends. The system is deployed through a Python-based microservice and integrated into a web-based academic platform via RESTful APIs for real-time inference. Experimental results using an 80:20 train–test split with 5-fold cross-validation demonstrate that the ensemble model outperforms individual classifiers, achieving 0.93 accuracy, 0.91 precision, 0.90 recall, 0.90 F1-score, and an AUC of 0.95. These findings confirm that the proposed framework provides a scalable, interpretable, and deployable solution that enhances predictive performance and supports early academic intervention.

Machine Learning, Random Forest, XGBoost, Academic Analytics, Ensemble Learning, Student Performance Prediction