Clustering, Classification, and Association Rule Mining for Educational Datasets

Abstract

Understanding students/pupils within the context of school is the focus of the expanding research topic known as educational data mining. Prior to the adoption of data mining tools in student analysis, it was challenging to identify students who were at risk of failure. However, using educational data mining tools has made it straightforward to examine student performance on previous exams. It helps teachers understand the kind of children they are working with. This knowledge can assist teachers to tailor their lecture notes to each student's needs and help difficult students focus more throughout the class. In this study, 1428 newly admitted student records from a tertiary institution in Nigeria were examined. Decision trees, association rules, and K-means clustering techniques were used to analyze the data. The results showed that arts courses were primarily responsible for low scores, social science courses for average scores, and the high JAMB scores came from science courses. Additionally, students from the geopolitical zones of the South-East, South-South, and North-East performed better than those from the geopolitical zones of the North-West, North Central, and South-West. The students with the highest JAMB marks were those who offered Physics and had an English score of at least 70. Most males offered Physics, Chemistry, Mathematics, and Biology while most females offered Economics, Government, CRS, and Literature in English. We highly recommend that Students who did not perform well should be given more attention and extra lessons/classes should be held for them to improve their grades in the various departments they were admitted.