Recursive clustering technique for students' performance evaluation in programming courses

Abstract

Automated prediction of students' performance in the earlier stage is a useful prospect in the teaching learning process. If the students who are probably going to fail are identified in the initial stage, a set of corrective measures can be taken to improve their grades. This paper employs the machine learning approach called the Recursive Clustering technique to group the students of the programming course into groups based on their performance in the prerequisite courses, co-requisite, CGPA and current course work result. Students present in the lower groups will be taken into consideration since they are highly prone to fail. Each of these groups will be provided with the set of programs and notes automatically based on their group. After a time period another assessment will be carried out and again the students will be clustered based on their new performance. This process will be repeated for three times so that most of the students from the lower group will move to the higher group. The results are compared to the number of students in each group before applying the recursive clustering technique and after. The results prove that this approach provides an effective way to predict the low performing students from their early in-class assessment, thereby enabling the student to be on track.