Hidden Markov Model-Based Approach for Adaptive Learning Path Recommendation

Adaptive learning systems require effective mechanisms to model students’ evolving learning states and provide personalized learning path recommendations. Traditional rule-based approaches are limited in capturing temporal learning dynamics and often fail to adapt to individual student progress. To address this limitation, this study proposes a Hidden Markov Model (HMM)-based framework for adaptive learning path recommendation. The proposed approach models students’ learning progression as a sequence of latent states inferred from discretized performance scores derived from learning activities. The HMM is trained using the Baum–Welch algorithm and employs the Viterbi algorithm to estimate the most probable sequence of learning states. These inferred states are then integrated into a recommendation engine that generates personalized next-item learning recommendations based on a probabilistic ranking mechanism and curriculum constraints. Experimental results on a real-world educational dataset demonstrate that the proposed model outperforms a traditional rule-based baseline, achieving an accuracy improvement of up to 15%. The model also shows improved robustness under varying learning behaviors and noisy data conditions. In addition, qualitative evaluation through learning satisfaction indicates positive user feedback toward the personalized recommendations. The findings suggest that the proposed HMM-based framework provides an interpretable and effective approach for modeling learning dynamics and generating adaptive learning paths. This work establishes a foundation for future extensions incorporating hybrid models, including deep learning and reinforcement learning, to further enhance adaptability and scalability in intelligent learning systems.