DENSITY BASED SPATIAL CLUSTERING MODEL FOR IDENTIFYING OUTLIER STUDENTS AND SLOW LEARNERS IN HIGHER EDUCATION

Abstract

Predicting student performance in higher education is an emerging topic because it directly correlates with educational quality. The categorization of students' knowledge levels enables educators to support them in improving their performance. Outlier students and slow learners are two significant student groups that require varying levels of assistance from teachers. The outliers are the students whose scores lie on both edges of the performance scale. They represent those who exhibit both extremely poor or outstanding learning abilities. Slow learners are students, who learn at a slower pace than their peers, but they do not necessarily fall into the failure category; they are below-average students. Cognitive assessment is a primary approach that allows educators to organize students based on their knowledge. The periodic class test or semester examination scores indicate the student's outcome in each subject. Subsequently, teachers can provide personalized instruction, employing different methods to teach the subject and supporting students to attain the best possible results. The contemporary advances in learning analytics and machine learning help to devise efficient models that predict the student's difficulties in advance. Existing studies utilized several parameters, including demographic data, study behavior, and academic details, to categorize the student group. However, this study aims to use academic scores to categorize students using K-Means and the DBSCAN algorithm using unlabelled data. The DBSCAN outperforms the K-means model with the highest silhouette and Davies-Bouldin index.