UNDERSTANDING COGNITIVE PROFILES OF STUDENTS IN A FIRST PROGRAMMING COURSE BY CLUSTERING THEM TO GUIDE PEDAGOGICAL INTERVENTIONS
This paper proposes a multidimensional diagnostic model concerning the understanding of cognitive profiles of students in computer science in a first programming course. To this end, we have clustered students based on the features from two different categories of data sources: (i) Data on students' grades obtained on the course of four related disciplines, considering data from the last six years and (ii) dataset based on source code regarding to problem-solving activities in programming. We then generate two sets of groups, G1 and G2, comparing them to generate the final clusters with the set of cognitive profile groups. As one result we have obtained five groups of cognitive profiles varying from students with no success to successfull. This result is in agreement with related work in the literature and brings more refined information. Furthermore, we introduce an interesting systematic way of generating groups of cognitive profiles, hoping to contibute to aid in decisions about pedagogical interventions in the context of virtual learning environments.