Research On KGCN Recommendation Algorithm Integrating User Information And Interest Evolution

With the rapid development of computer technology,the popularity of the internet and the rise of e-commerce,users are faced with a huge amount of data and have difficulty finding the information they need.As a result,recommendation systems have emerged.Recommendation algorithms are the core of recommendation systems and have become a hot research topic in the field.Collaborative filtering(CF)algorithm is a traditional recommendation algorithm that mines and analyzes data based on users’ historical behavior records to discover their interests and hobbies for recommendation.However,this algorithm has problems with poor interpretability,data sparsity,and cold start.These issues are generally solved through mixed recommendations with auxiliary information,especially knowledge graphs,which can effectively solve such problems.Based on this,this paper proposes a KGCN recommendation model that integrates user information and interest evolution and applies it to a movie recommendation system,effectively solving the problem of information overload faced by movie websites.The main contents of this paper are as follows:Analyzing the research status of traditional recommendation algorithms such as CF and knowledge graph-based recommendation algorithms,summarizing the characteristics and shortcomings of various algorithms.Constructing a knowledge graph in the movie domain,based on the Movielens dataset,by extracting the entity relationship of movie data in the dataset,building entity-relationship triples and storing them in the Neo4 j graph database to construct a movie knowledge graph.In view of the shortcomings of the KGCN recommendation model in only considering project content and ignoring user information features in the information recommendation process,this paper builds a KGCN recommendation model that integrates user information and interest evolution(Research on Knowledge Graph Convolutional Networks Recommendation Algorithm Integrating User Information and Interest Evolution,KGCNUIE)based on KGCN.The recommendation model includes three modules: data processing,interest evolution,and model prediction.The data processing module prepares the user’s long and short-term rating dataset,builds a corresponding knowledge graph,and integrates the user’s personal information into the initial embedding vector.The interest evolution module models the user’s long-term interest preference and short-term preference using attention mechanisms.The model prediction module generates user interest representation by combining long and short-term interest representation and performs rating prediction with the target project.This paper uses Movie Lens-1M and Book-Crossings as experimental datasets and compares and tests the model with multiple evaluation indicators,verifying the effectiveness of the algorithm.A movie recommendation system is built based on the KGCNUIE model as the recommendation algorithm to solve the problem of resource overload on movie websites.Through interface design,database design,and movie recommendation design,a movie recommendation system based on the KGCNUIE model is realized.Through testing and analysis of the recommendation system,the effectiveness of the system is verified.