An AI-Driven Framework for Enhancing Students’ Learning Efficiency Using a Performance Evaluation System

Learning environments in higher education have become more complex, and the range of student needs has widened. This has exposed the weaknesses of the traditional performance evaluation methods that many institutions still use. Manual assessment is often slow, hard to scale, and not personalized, and these gaps limit students’ learning efficiency. This paper presents an AI-driven framework, called the AI Performance Evaluation System , that uses machine learning and learning analytics to study student engagement, performance patterns, and behavior, and then gives personalized feedback and adaptive learning paths. The system was built with modern web technologies: React.js for the frontend, Node.js and Express.js for the backend, Supabase for the database and authentication, and a FastAPI microservice for model inference. A Design Science Research methodology together with a mixed-methods approach was used. The models were trained and evaluated on a public student-performance dataset of 1,044 records (395 Mathematics and 649 Portuguese records) using five-fold stratified cross-validation. For identifying at-risk students, logistic regression gave the best and most stable results, with 90.4% accuracy, 0.92 recall, 0.81 F1-score, and 0.97 ROC-AUC, clearly above the 78% majority-class baseline. A linear regression model predicted the final score with an R² of 0.83 and an average error of about 1.6 points on a 20-point scale. K-means clustering separated learners into three behavior profiles that were used to shape recommendations. The classifier degraded gracefully when up to 20% of the input values were missing, and a subgroup check by gender and location showed only small performance gaps. Pilot users reported that the dashboards and personalized feedback were clear and useful. The study adds a single framework that joins classification, regression, clustering, and text analysis for performance evaluation, and it discusses privacy, security, and bias for use in higher education, with attention to developing-country settings.