Explainable Recommendation System Based On Dynamic Knowledge Graph And Reinforcement Learning

In today’s society,the recommendation system has penetrated into all aspects of life.Every Internet application used in daily life usually uses the recommendation system as a key part to provide users with items that meet their preferences.However,as life becomes more and more bound to the Internet,the past purely algorithm is no longer sufficient to meet the needs of society.Regarding the existing recommendation system,even if the administrator finds that there are problems with the recommendation results,such as recommending inappropriate videos or products to minors,due to the black box characteristics of the recommendation system,it is difficult for the administrator to effectively perform the system adjustment.Therefore,it is extremely important for companies that are paying more and more attention to social responsibility to fully understand the recommendation process and realize the explainability of the recommendation system.In order to solve the problem of insufficient explainability faced by the current recommendation system,this paper studies the explainable recommendation system,and determines to use the most advanced knowledge graph and reinforcement learning to build the overall architecture of the recommendation system.At the same time,in order to further improve the practicability of the system and solve the problem that the static knowledge graph embedding model consumes a lot of time and resources in the update process,this paper studies the dynamic knowledge graph embedding model and applies the dynamic knowledge graph embedding model Dynamic Knowledge Graph Embedding(DKGE)to the final design and realized a set of explainable recommendation system based on dynamic knowledge graph and reinforcement learning.The system lets the reinforcement learning agent to search in the knowledge graph,taking the final searched target node as the user’s recommendation result,and the search path as the reasoning path of the recommendation result.In order to improve the accuracy of the recommendation,the Beam Search algorithm is used to select the path with the highest probability among multiple search paths as the recommendation result.At the same time,in order to facilitate users to use and interact with the recommendation system,as well as to view the knowledge graph structure and information,this article takes the recommendation system as the core and builds a website system to provide a convenient interface for users to view and use.Through the research and analysis of the knowledge graph and recommendation system,the system uses the recommendation system architecture based on knowledge graph and reinforcement learning to solve the problem of poor explainability of the current recommendation system;the dynamic knowledge graph embedding model solves the problem that It is difficult to update the knowledge graph during the application process.The recommendation system that meets the needs of enterprises can be explained,as well as the need to save time and resources in the training process.