Analysis of student achievement scores via cluster analysis

Abstract

In education, the overall performance of every student is an important issue when assessing the quality of teaching. However, in the traditional educational system not all students have the same opportu nity to develop their academic skills in an efficient way. Different teaching techniques have been proposed to adapt the learning process to the stu dent profile. In this work, we analyze the profile of students according to their performance on academic activities and taking into account two dif ferent evaluation systems: work-based assessment and knowledge-based assessment. To this aim, data was collected during the fall semester of 2019 from a physics course at Universidad Loyola Andaluc´ıa, in Seville, Spain. In order to study the student profiles, a clustering approach com bined with supervised feature selection was applied. Results suggest that two student profiles are clearly distinguished according to their perfor mance in the course in both evaluation approaches. These two profiles correspond to students that pass and fail the course. The output of the analysis also indicates that there are redundant and/or irrelevant fea tures. Therefore, machine learning techniques may be helpful for the design of effective activities to enhance the student learning process in this physics course.