Research And Application Of Movie Recommendation Model Based On Knowledge Graph

With the rapid development of the film industry,the Internet has been regarded as an important channel for the public to obtain film and television information.However,information overload is a common problem.The personalized movie recommendation can support users to discover the target movie from massive amounts of movie information by filtering the redundant data and obtaining the high-quality movie information.The traditional recommendation algorithm still has great potentials for making further progress in the data sparsity and the cold start.Therefore,improving the above problems by some auxiliary information of the recommendation algorithm becomes the main trend of recommender systems research at present.As an emerging auxiliary method,the knowledge graph provides a new way for the improvement of the recommendation system.Therefore,for the field of movies,the movie recommendation system based on the knowledge graph auxiliary is analyzed and studied in this paper.The main research contents are as follows.(1)Extract the webpage data from the Douban Movies website,and construct a knowledge graph of the movie domain.In this paper,the movies data set is constructed by utilizing the web crawler tool,which achieves the division of the entities,relationships and attributes in the movie field.The data is imported into the Neo4 j graph database,which completes the construction of the movie knowledge graph.(2)A movie recommendation model CMKR based on the knowledge graph is proposed.By means of a two-stage multi-task learning framework and the combination of the convolutional neural network and the MKR recommendation model,the model uses the convolutional neural network to optimize the initial value of the vector to dig out more user and movie characteristic information,and alleviating the data sparse problem.Then,the interactive data of the user and movie is obtained by adopting the multi-task learning framework and the alternate training method and combining the knowledge graph and the recommendation system.Finally,the similarity between the users and movies is calculated according to the feature data,and the recommendation score of movies without interactive recording is predicted,thereby reducing the system cold start problem.The experimental analysis shows that the CMKR model has a good performance in the movie recommendation,which is better than the comparison models in the various recommended evaluation index.(3)Finally,according to the demand analysis of the movie recommendation model,a recommendation prototype system based on the movie knowledge graph is designed to further verify the model proposed in this paper.Based on the constructed movie knowledge graph data,the movie recommendation system utilizing the CMKR recommendation model to provide personalized movie recommendation for users.