Frequent pattern mining augmented by social network parameters for measuring graduation and dropout time factors: A case study on a production engineering course

Identifying factors that lead students either not to graduate on time or drop out helps Higher Education Ina stitutions improve retention and decrease attrition. This paper tackles this problem by introducing a novel approach to discovering such factors through pattern mining using association rules. The novelty of the method arises from introducing social network analysis inside the pattern mining process. The social networks metrics for each student and the degree of propagation of grade point average are computed and integrated with students’ records for pattern mining serving as a proxy for the existing bond among students, which is a relevant factor for attrition and dropout analysis. This paper examines the Bachelor Program in Production Engineering at the Federal Center for Technological Education of Rio de Janeiro from 2011 to 2017. Our experiments indicate congruence with the literature: (i) lower school performance leading to delay; (ii) higher performance leading to graduation in optimal time. Besides, our new method sheds light on students with little participation in the social network who are more likely to delay or drop out. Our findings may aid managers in discovering students with patterns that can indicate imminent lateness or dropout.