Research On Recommendation Algorithm Based On Item Cooccurrence And Knowledge Graph

With the rapid development of the Internet industry and information technology,users have been provided with a large number of information options.Therefore,the recommendation system helps users deal with the problem of information overload,which has received more and more attention.Faced with such a large-scale service,users will only call a small part of the service,resulting in data sparsity.Therefore,in the case of sparse data,how to establish the relationship between features and mine more data information to better help users filter redundant information and meet users' personalized needs has become a huge challenge in today's society.The traditional collaborative filtering algorithm has achieved good results in the recommendation system,but when encountering data sparse problems,its algorithm performance will be hindered.In order to solve the problem of data sparsity,this paper proposes a novel recommendation model,namely,metric factorization with item cooccurrence for recommendation(MFIC),which uses the Euclidean distance to jointly decompose the user-item interaction matrix and the item-item cooccurrence with shared latent factors.The item cooccurrence matrix is obtained from the colike matrix through the calculation of pointwise mutual information.At the same time,experiments on rank prediction and item ranking were performed on 4 data sets.Experimental results show that the method is significantly better than the comparison algorithm in both evaluation and prediction and item ranking.In addition,it is not only suitable for rank prediction and item ranking,but also can well overcome the problem of sparse data.However,the complexity of the model has increased,which needs to be effectively mitigated in the follow-up study.Knowledge graph recommendations that the rich facts and connections in the knowledge graph are used as user-item supplementary information to address data sparseness and cold-start issues.In order to make full use of the advantages of knowledge graph,in this paper,we propose a recommendation algorithm for propagating user preference knowledge graph(PUPKG),an end-to-end recommendation framework based on embedded and path-based knowledge graph recommendation is mainly used.At the same time,Manhattan distance and matrix decomposition are introduced to train the potential factors of users and items together to alleviate the problem that the matrix inner product does not meet the triangle inequality theorem.In order to show the effectiveness of our method,extensive experiments have been carried out in the data sets BookCrossing and Movielens-1M.The experimental results show that the method has a great improvement in both datasets.However,we only use some common relationships to build knowledge graph.In the future,we should introduce new relationships to improve the establishment of knowledge graph.