AI-Based Student Performance Prediction Using Machine Learning Techniques

Predicting student performance is an important application of Artificial Intelligence (AI) in education. Educational institutions generate large amounts of data, including examination results, attendance records, assignment scores, and Learning Management System (LMS) data. Machine Learning (ML) techniques can be applied to these datasets to predict academic achievement and provide educators with valuable insights into identifying students who may benefit from timely academic interventions. This review provides an overview of common student performance prediction tasks, educational data features, machine learning techniques, evaluation methods, and Explainable Artificial Intelligence (XAI). The literature covers various prediction tasks, including classification to predict academic outcomes (such as pass/fail status or grade categories), regression to estimate numerical scores, and early-warning systems to identify students at academic risk. The machine learning techniques explored include Decision Trees, Random Forests, Logistic Regression, Naive Bayes, Support Vector Machines, and Neural Networks. Due to differences in datasets, prediction objectives, selected features, and validation methods across existing studies, comparing model performance remains challenging. Key issues include limited generalisability across institutions, class imbalance, model interpretability, fairness, data privacy, and insufficient evidence regarding the effectiveness of recommended interventions. This review highlights these challenges and proposes a responsible conceptual framework covering data preparation, model comparison, appropriate evaluation, explainability, and educator oversight. The proposed framework provides a foundation for future research into reliable, transparent, and responsible machine learning applications for student performance prediction.