Predictive models to enhance learning based on student profiles derived from cognitive and social constructs

Abstract

A preliminary exploratory and predictive model to correlate the academic performance of a sample of 96 students enrolled in different basic engineering courses with cognitive and social constructs is presented. The model integrates several dimensions regarding Multiple Intelligences, Self-Regulation skills and Learning Styles constructs. The exploratory study is carried out with three statistical methods: analysis of principal components, correlation analysis and cluster formation. The prediction of students' final grades was accomplished from three perspectives: i) from the average final grade in each cluster, ii) obtaining rules to classify, a-priori, each student as “pass” or “fail” by means of decision trees, and iii) detecting those dimensions of the constructs that have a larger impact on students' grades, using linear regressions. It is found that the logical-mathematical intelligence has the largest positive impact and the anxiety of the students also has a significant, but negative, impact. It is also found that students who present a high intrinsic motivation are very likely to pass their courses. Additionally, it is found that the average grades in each cluster are the expected ones according to the characteristics defining the cluster. The results are encouraging and may serve to improve instructional design and the elaboration of more tailored didactic resources.