RESEARCH ON MACHINE LEARNING METHODS FOR PERSONALIZED LEARNING RECOMMENDATION SYSTEMS

The present study proposes a personalised recommendation system for students, the function of which is to recommend teaching resources according to students' mastery and interest. Existing information platforms are challenging to utilise for the purpose of recommending teaching resources that align with students' individual conditions, and conventional recommendation technologies frequently encounter issues with accuracy and adaptability. It is evident that a combination of deep neural networks, reinforcement learning mechanisms and domain knowledge graphs is required. Firstly, deep learning methods are employed to extract multi-dimensional features from learning trajectories. Subsequently, reinforcement learning frameworks are utilised to optimise the recommendation decisions in order to facilitate dynamic planning of learning paths. The establishment of a structured knowledge graph enables the system to establish semantic associations between learning resources with a high degree of effectiveness, thereby improving the accuracy of recommendation results. In the face of the challenging issue posed by the limited number of new users, the system has been engineered to ensure the consistency of its recommendation quality in the majority of instances where the initial data is found to be inadequate. This is achieved through the implementation of a domain migration learning scheme.