Predicting student progression as a Markov chain

Abstract

 Effective school management may benefit from accurate student progression models to optimize the allocation of instructors and teaching spaces. However, existing research predominantly focuses on Markovian models for year-based programs, leaving a gap in the literature for credit-based undergraduate courses. This paper proposes a framework for modeling the progression of a group of students in credit-based programs using Markov chains, without requiring extensive training. The research addresses three key questions: (1) Can student enrollment and progression be modeled as a Markov chain? (2) Does a multi-agent simulation of a Markov model accurately replicate student-group behavior? (3) Are the multiagent Markov model simulation results akin to real student progression data? To answer these questions, we analyzed student progression data from an engineering program to extract probability distributions for course failure, class enrollment, and dropout rates. These distributions were then used to develop Markov chain models, which were simulated using a multi-agent approach and were compared the real data progression rates. Our findings indicate that a Markov chain model can effectively emulate student progression in engineering programs, and the results from our multi-agent simulation closely match the actual progression data. This framework provides a valuable tool for school management and course demand prediction in credit-based undergraduate programs.