| Recommendation algorithm is one of the most effective tools to solve the problem of information overload,and has been widely researched by many scholars both here and abroad.Based on the characteristics of dimensionality reduction,nonnegativity and strong explanatory power,in this paper,we will use non-negative matrix factorization as modeling and calculation tools,and in-depth study the recommendation algorithm,in order to improve the accuracy of recommendation algorithm.Firstly,this paper analyzes the research background,significance and present current situation of the recommendation algorithm,then in order to compare the recommend effect of different algorithms,the paper give the evaluation criteria of the recommendation algorithm.Secondly,this paper proposes a recommendation algorithm based on directed graph partition.The algorithm combine bipartite graphs network structure with resource allocation method,and establish the directed graph of relationships between objects,as well as use asymmetric nonnegative matrix decomposition method to split the directed graph,and finally realizes the recommendation from the Top-N items.Experimental results show that the proposed algorithm can improve the recommendation accuracy.Thirdly,this paper proposes a algorithm by making use of JNMF that combine the collaborative filtering based on user and collaborative filtering based on item.Because of the joint nonnegative matrix factorization can reveal the relationship of the complex network structure which can combine the advantages of the two algorithms,it can reduce mean absolute error and improve the accuracy of recommendation.Fourthly,in order to solve the problem of sparse data in recommendation algorithm,this paper proposes a new algorithm to solve the problem of sparse data.In the similarity calculation,the user attributes and the item attributes are incorporated to improve the accuracy of similarity calculation.In addition,the Hadamard product is introduced into the objective function of the algorithm to solve the problem of the scoring data sparseness in the recommendation algorithm.Finally,the research work of this paper is summarized and the future work is prospected. |
