Smart Academic Insights: Predicting Student Performance with Machine Learning-Driven Result Management

Abstract

The prediction of student performance plays a significant role in improving academic outcomes and
providing effective, personalized guidance to students. Traditional research methods typically focus on
identifying common characteristics shared by groups of students to understand general learning trends.
While this approach provides valuable insights, it often overlooks the unique differences in individual
learning patterns, which are essential for delivering tailored educational interventions. Addressing this
gap requires methods capable of capturing both group-level patterns and individual distinctions in
learning behavior. To enhance the segmentation of multidimensional discrete data, a novel framework has
been proposed that integrates a relationship matrix-based bipartite network (RMBN) with Louvain
clustering. This innovative technique facilitates more accurate and meaningful clustering of students based
on various academic attributes, enabling the identification of specific learning behaviors and challenges.
By effectively grouping students with similar patterns while preserving essential individual differences, this
method enhances the ability to draw precise and actionable insights from complex educational datasets. In
addition to the clustering framework, the study also introduces a hybrid neural network model (RMHNN)
designed to overcome the limitations faced by traditional algorithms when processing discrete and diverse
data types. This model, when applied to real-world student datasets, demonstrated exceptional
performance, achieving a prediction accuracy of 93.1% and an F1-score of 90.45%