Recommendation Algorithm based on Graph Network Representation Learning

Graph representation learning has the property of causal inference of relation, and integrates graph structure with recommendation system, which effectively reduces the number of iterations required for parameter optimization. Using the method of graph representation learning to recommend can better deal with topological associations and semantic features in graph structure, and effectively enhance the accuracy of recommendation. Firstly, the neural cooperative filtering algorithm is studied to simulate the interaction between a user and learning resources with multiple perceptron, so as to better model the interaction between complex users and learning resources. Then, aiming at the uncertainty of user preference and the ambiguity of description, the short-term preference model and long-term preference model are constructed. Finally, the experimental environment is constructed, the experiment is carried out with the Amazon books data set, and the experimental results are compared and analyzed, which verifies the superiority of the proposed algorithm.