So sánh hiệu năng các mô hình học máy trong dự báo kết quả học tập sinh viên ngành Công nghệ thông tin và Kỹ thuật phần mềm, Trường Đại học An Giang

Abtract

Predicting student academic performance is a critical application of artificial intelligence in education, enabling personalized learning pathways and early intervention for students at risk of academic difficulties. This study evaluates and compares the performance of nine machine learning models, including basic algorithms (KNN, SVM, DCT, MLP) and ensemble models (RF, Bagging, AdaBoost, XGBoost, CatBoost), in forecasting the graduation rankings of students in the Information Technology (IT) and Software Engineering (SE) programs at An Giang University. The dataset, comprising 788 student records from graduates between 2019 and 2024, was preprocessed, balanced, and standardized. Results indicate that SVM achieved the highest performance on the IT dataset, while CatBoost outperformed others on the SE dataset. The study confirms that selecting an appropriate machine learning model should be based on the specific characteristics of each program's data, highlighting the potential of these models for integration into academic advising and early warning systems in higher education institutions.

Keyword: Machine learning, academic performance prediction, educational data mining, Information Technology, Software Engineering

Download