Modeling academic performance using subspace clustering algorithm

Abstract

Reforms in the educational system focuses more on continuous assessment. Student performance study has been the primary challenge for any course having continuous assessment to validate whether the course objectives are met and also to identify the areas of the course structure that needs improvement. This paper analyses the student performance in different core subjects of the course with diverse types of competency components such as presentation, assignment, quiz, case-study etc. along with written examination in order to test the knowledge of students as well as their interest in the subject. PROCLUS algorithm has been considered for experimentation as the algorithm identifies similarities among the data sets and forms subspace clusters. The algorithm not only considers random sample points, but also successfully scans the entire data set to identify meaningful dimensions that are needed to form actual clusters. The experimental results prove the effectiveness of Proclus algorithm in categorizing the students as per their competency as well as predicting their future performances. The quality of clusters formed with PROCLUS has been proved to be of high dimension that finds patterns across the students’ performance.