Research on the Design of Recommendation System for Learning Methods Based on Bayesian Networks

The current mainstream machine learning algorithmic recommender systems suffer from the problems of difficulty in training at low feature counts, probabilistic operations, and inaccurate recommendation results derived in data sparse environments. To address these problems, this study builds a learning path recommendation model based on Bayesian networks that can use Bayesian network models to characterize the relationship between knowledge points, diagnose learners' mastery of a certain knowledge point by monitoring learners' learning behaviors and learners' tests, and adjust the learning content scheduling scheme according to learners' differences. The experimental results show that the overall hit rate, normalized loss gain, and coverage rate of the Bayesian network-based learning path recommendation model are higher than those of other algorithmic models, with the overall hit rate above 61%, the normalized loss gain of 39.3%, and the coverage rate of 41.8%, and the Bayesian networkbased learning recommender system is still able to give personalized recommendations when the learner profile information is small. The results of this research have some value in the field of recommender system development and can be used as a technical reference.