Investigating learning outcomes in engineering education with data mining

Abstract

AbstractHigher education institutions are catching up on their high competition and challenges are in their analysis productivity. The major challenge is to monitor and analyze student progress through learning outcomes in the curriculum. One of the approaches is the outcome‐based education (OBE) model to deal with learning outcomes. OBE is an integral part of higher education institutions. The OBE system is a key step for accreditation in engineering education. OBE focuses on a student‐centered approach. The OBE is not restricted to well‐defined teaching strategies or direct evaluations but also encompasses indirect evaluations to help students achieve the intended outcomes. In this investigation, engineering students’ data have been analyzed forming three distinct clusters to group students according to best, average, and worst achievement of learning outcomes in two different computer engineering courses generally taught in the early semesters in higher education institutions. A data mining clustering approach is used to segment students using k‐means and k‐medoids techniques. Clustering can be regarded as a data modeling technique that provides summary data that interact with multiple disciplines and plays an important role in a wide range of computer applications. The investigation comprises of two parts for analysis: one part of the analysis is the mid‐term and final exam scores, the quiz and assignment results, the laboratory results, and the evaluation, together with the learning outcomes achieved, and the other part is the comparative analysis of learning outcomes achieved in both engineering courses clustering with the best, average, and worst attainments, respectively. In this investigation, the results obtained from clustering data points show that the same group of clusters with the best, average, and worst learning outcomes achievements formed using both k‐means and k‐medoid clustering for one course. On the other hand, a diverse group of clusters with the best, average, and worst learning outcomes achievements formed using both k‐means and k‐medoids clustering for another course.