Research On Explainable Recommendation Model Based On Knowledge Graph

With the development of information technology and the Internet,people have entered the era of information explosion,where massive information is generated at every moment.In this context,recommendation systems have been proposed to solve the problem of information overload and to provide recommendation results that meet the needs of different users according to their interests.Recently,scholars have gradually focused their research on providing explanatory basis for the user while generating recommendation results,telling the user why the item is recommended to him.Therefore,in this paper,we focus on recommending the desired items for users in a more acceptable way and generating corresponding explanations.In order to reach the goal of improving the probability of users choosing recommended items and the recommendation performance,knowledge graph is introduced into explainable recommendation systems,explainable recommendation models are constructed,optimization algorithms are employed to solve the constructed models.The main researches of this paper are as follows.(1)In order to achieve the purpose of improving the explainability of recommendations while maintaining accuracy and diversity,knowledge graph is constructed,and the explainability of recommended items is quantified according to the similarity between entities in the constructed knowledge graph in terms of attributes,then an explainable recommendation model is constructed and multiobjective optimization algorithm is used to solve the model,finally the best path between the target user and the recommended items in the constructed knowledge graph is selected as the basis for explanation.(2)Path-based approach only considers the entities in knowledge graph but ignores the relationships among entities,and embedding-based approach is employed to measure the importance of relationships to entities.In order to fully use the entities and relationships in knowledge graph to obtain better recommendation results,unified-based approach is used to quantify the explainability of recommended items,an explainable recommendation model is built,many-objective optimization algorithm is used to simultaneously optimize accuracy,diversity,novelty,and explainability.Finally,the best path between the target user and the recommended item is selected as the explanation in constructed knowledge graph.(3)An explicable recommendation model is constructed based on preferences of user,which first obtains the average rating of target user on all items,and the average rating is used to reflect overall preferences of the user.A pruning strategy is designed to delete or merge entities with little significance and their corresponding paths to streamline constructed knowledge graph network structure based on preferences of user.Paths are used as decision variables to improve the crossover operator and variation operator on the basis of the analyzed path features,and the best recommendation paths and corresponding recommended items are obtained by many-objective optimization algorithm.