Recommendation System Based On User Multi Interest Knowledge Graph

In recent years,with the application and development of mobile Internet,network data is growing explosively,and information overload is becoming a very serious problem.Facing the massive data on the Internet,people cannot quickly and clearly distinguish between miscellaneous data and interesting modules.The recommendation system is an effective information filtering tool which suggests items or services that users may be interested in.It exists to provide access to a greater variety of effective information and services,help the user to quickly and accurately discover useful information and to be able to present information that best matches the user’s interests.As a popular application in the Internet field,recommendation systems are widely used not only in social and ecommerce,but also in movies,music,books and articles.The recommendation quality of traditional recommender systems is not high,mainly due to the problem of data sparsity and cold start.By introducing auxiliary information into the recommender system,the recommendation effect can be well enhanced,such as context information,user or item attribute characteristics.When the recommendation system is introduced,the sparsity and cold start problems of traditional recommendation systems can be solved.As a heterogeneous information graph,knowledge graph is an effective way to effectively integrate various auxiliary information and enhance the performance of a recommendation system.Although the existing recommendation system based on knowledge graph has achieved good results in result prediction,due to knowledge graphs are only augmented as auxiliary information,which cannot capture changes in users’ interests and are difficult to deal with multimodal data forms.In order to solve the above problems,we conducted research on the representation of knowledge graphs and the application of knowledge graph in recommendation system.The contributions are concluded as follows:(1)In order to provide more accurate,diverse and explanatory recommendations,a recommendation system was proposed based on the users’ Multiple Interest Knowledge Graph(MIKG).By introducing the user interest knowledge graph,the user interest is explicitly represented as a knowledge graph entity,and the knowledge graph embedding is used to represent the users’ interest knowledge graph.At the same time,a multi-interest module is introduced to connect the recommendation module and the knowledge graph embedding module to jointly learn item and entity features.And diversified processing of users’ interests,so as to enhance the semantic information of items,so that the recommendation system can better meet people’s needs for personalized service.(2)According to the practical application scenarios based on the user interest knowledge graph and the requirements of the recommendation system applicable to multiple scenarios,the prototype article recommendation system based on the user multi interest knowledge graph was designed and implemented.The multi-modal data was unified into the recommendation system model by associating the multi-modal information with the user interest knowledge graph.And a recommendation system that conforms to multiple scenario recommendation was built.