Multi-layer Interest Modeling Combined With Knowledge Graph And Its Application In Recommendation System

The rapid development of science and technology is accompanied by the explosive growth of data volume,and "information overload" has become one of the problems that people need to solve urgently.As a currently widely used information filtering method,the personalized recommendation system actively recommends items of interest to users,greatly reducing the time cost of decision-making,and enabling people to obtain rich and diverse information services.As a kind of multi-source heterogeneous information,the knowledge graph can provide rich prior knowledge for recommendation algorithms,fine-grained characterize the semantic associations between users or items,effectively solve the problem of data sparsity,and improve the performance of the recommendation system.However,the numerous existing algorithms cannot find the deep interest of users only by relying on historical interactions,and the cold-start problem of new users has always hindered the recommendation system.Aiming at the above problems,this dissertation proposes a new user interest model,and uses it as algorithm support to design the movie recommendation system.The specific research content is as follows:(1)This dissertation proposes a multilayer interest model(MIKU)that combines knowledge graphs and user attributes.The model first uses the user's historical interaction items as the head entity of the knowledge graph to construct shallow interests,and combines the relational paths in the knowledge graph to link the related entities of the historical items to mine the user's deep interests;secondly,considering the diversity of user interests,adaptive weighting mechanisms are adopted for different levels of interest to learn user preference weights for different behaviors and points of interest.While the fine-grained characterize the item features,it also more comprehensively characterizes the user's interest,and combines the user's attribute characteristics to effectively solve the cold-start problem.Validation on the public movie data set shows that compared with many benchmark models,the MIKU model improves the accuracy of recommendation results by 1.93%?5.59% and the recall rate by2.95%?4.7%.(2)In order to verify the feasibility of the proposed algorithm in the real world,this paper uses Douban website to crawl and process real data information,build a complete Douban movie knowledge graph based on related technologies,and design and implement a movie recommendation based on the knowledge graph system.First,the system requirements are analyzed based on the principle of user experience,and the overall is designed framework from the three modules of data acquisition,knowledge graph construction and movie recommendation system implementation,and analyze the design methods of each module in detail.The specific implementation of the system adopts the Python language,uses crawler technology and PyQt5 integrated library to complete the development of a movie recommendation system based on the knowledge graph,and combines the Neo4 j graph database to store and visualize the knowledge.Through detailed functional testing and interface optimization of the system,while ensuring user needs,the knowledge graph can be used to provide a certain interpretability for the recommendation list to ensure the feasibility of the system.