Evaluating Machine Learning Techniques for Student Performance Prediction in Higher Education

Accurate prediction of a student’s performance is beneficial in improving learning outcomes, while the conventional machine learning models are unable to reflect the progressive and pyramidal structure of learning. This paper presents Hierarchical Adaptive Neural Networks (HANNs), a new algorithm that uses hierarchical learning in its architecture as well as modifies the architecture to better predict student performance. HANN utilizes a two-layer approach: One for student records and another for collective learning patterns capable of changing its neural structure at a given period in relation to the existing patterns. Precise evaluation of HANN over ten benchmarks of higher education datasets showcases the better performance of the proposed HANN model in comparison to the traditional models.