Modeling Engineering Undergraduate Student Success using K-means Cluster Analysis

摘要

Early prediction of engineering undergraduate academic achievement would be valuable for both faculty and administration. Prior student performance and other psychosocial factors can provide valuable insights. Typically, such non-cognitive factors are not well-studied in academia. Using modern clustering tools, such as K-means clustering, this study attempts to reveal hidden patterns and map students based on a combination of their prior knowledge and a self-rating of their attitudes and skill-related behaviors. In this study, approximately 320 students responded to a questionnaire. Statistical analysis using SPSS 24 found three clusters, i.e., high performing students (25.2%), average students (44.3%), and low performing or at-risk students (30.4%), based on two cognitive and twelve non-cognitive variables. Grouping students using K-means clustering can provide educators with valuable insights that can benefit administrators, academic advisors, and instructors relative to academic planning and resource allocation. Based on the results, the study proposes the provision of more tailored support services and programs for each cluster to further enhance student success and prevent any immature failure.