Predicting Student Academic Success with Hidden Markov Models

Abstract

This research paper presents an approach for predicting student academic success using Hidden Markov Models (HMMs). Leveraging a comprehensive dataset encompassing students' demographics, academic performance, attendance records, and course engagement, the study employs an HMM framework to model levels of student academic success. Observable emissions derived from the data, such as grades and interaction patterns, are utilized to train the HMM and infer the most likely sequence of hidden states for new students. Evaluation of the proposed model demonstrates promising predictive accuracy. Through rigorous assessment using standard metrics including state prediction accuracy and state transition accuracy, the effectiveness of the HMM in capturing diverse student trajectories is demonstrated, underscoring the potential of HMMs as a powerful tool for understanding and predicting student outcomes, offering valuable insights for educational interventions and support systems.