A fuzzy logic-driven machine learning framework for multi-class classification of students academic performance

The foundation of a nation's future lies in the performance of its students. Their academic success shapes not only their personal growth but also the strength and progress of the entire country. High-achieving students go on to become skilled professionals, innovators, and leaders who drive economic development and social change. In this way, student performance plays a vital role in building a knowledgeable and competitive society. Investing in education today ensures a stronger, brighter nation tomorrow. Student performance depends on various factors such as the quality of teaching, availability of learning resources, parental support, and a conducive learning environment. Motivation, mental health, and socio-economic background also significantly influence how well a student performs academically. Although previous studies have applied either clustering techniques or machine learning algorithms independently to evaluate student performance, they often lacked the ability to capture the uncertainty and overlap in student characteristics. This research addresses that gap by combining fuzzy c-means clustering, which allows for soft classification, with a machine learning algorithm. The hybrid model enhances the accuracy and interpretability of performance evaluation by leveraging the strengths of both methods. This integrated approach provides a more nuanced and data-driven understanding of student outcomes. In addition to this, explainable artificial intelligence technique is employed to provide a transparent and interpretable summary of how the proposed hybrid model functions. By integrating this technique, the research not only improves prediction accuracy but also ensures that the decision-making process is understandable to educators and stakeholders. Furthermore, feature distribution graphs and correlation heatmap are drawn to have visual understanding of the related features. At the last, comparison with existing techniques, limitations and future directions are being discussed.