Mining direct acyclic graphs to find frequent substructures — An experimental analysis on educational data

Abstract

The number of undergraduate students joining universities in Brazil has largely grown in the recent years. However, the number of students who actually graduate remains low. Some studies show that this is due to a phenomenon called retention, consisting of a student taking more time to graduate than the minimum required by the program, which may lead to late graduation. Hence, identifying retention patterns in an undergraduate program may assist the universities in anticipating the entrance of qualified professionals in the job market, while lessening the students’ dropout rate. Undergraduate programs and grade reports can be represented by DAGs, in which each course (as a task to be accomplished by each student) is represented as a vertex, and relations between courses are represented as edges. This article proposes methods for mining DAGs using statistical analysis and Apriori-based concepts, to identify retention patterns in undergraduate programs. This work also presents an experimental analysis using real data from Fluminense Federal University, a Brazilian public higher education institution, for evaluating the methods.