Research On Data Sparsity Of Collaborative Filtering Recommendation Algorithm

With the continuous development of information technology,The era of big data came quietly,accompanied by the information overload problem for the Internet age which is undoubtedly a huge waste of time and resources.Personalized recommendation algorithm is an effective solution to this problem and provides great convenience for people's choice.Collaborative filtering recommendation algorithm has the characteristics of simple,easy to implement,and the novelty of the recommendation.These make it become one of the most successful techniques in personalized recommendation algorithms.However,when the number of users and projects is increasing,and the number of user operations on the project does not substantially increase the situation,then the user-project data may become sparse.This reduces the accuracy of the neighborhood search required by the cooperative filtering recommendation algorithm,and the number of nearest neighbors is too small,which has important influence on the recommendation quality and recommendation accuracy of the whole recommendation system.In this paper,we study the important effect of data sparsity on collaborative filtering recommendation algorithm,and offer the following solutions:Firstly,we propose a collaborative filtering recommendation algorithm based on singular value decomposition(SVD)and fuzzy clustering.By reducing the number of eigen values in the special relativity theory of physics,the dimensionality of dimensionality reduction can be determined more accurately and the dimensionality reduction of the original data and its data filling can be realized.In addition,using the method of fuzzy clustering to cluster the similar users,so as to achieve the purpose of reducing the search range of the neighbor users.Experimental results show that the proposed algorithm can alleviate the impact of data sparsity,and improve the quality and accuracy of the system.Secondly,by combining the trust relationship between users and the score similarity of the project,we design a collaborative filtering recommendation algorithm based on fusion trust user.The operation trust relationship between the users constructed by using directed network graphs compensates the shortcomings that the relationship between users can not be accurately measured by merely calculating the similarity between users.Experimental results show that the algorithm can reduce the data sparsity effect to a certain extent and improve the quality and accuracy of the system.Finally,on the basis of the first two methods,we conceive a way to improve the credibility.In order to solve the validity problem of the neighbor selection,we propose a collaborative filtering recommendation algorithm to improve the credibility.The SVD dimension reduction is combined with the trust relationship among users to improve the credibility of neighbor selection.Furthermore,by means of the project itself and the subjective forecast of the users to improve the credibility of the final prediction score.Experimental results show that the proposed algorithm can improve the quality of recommendation while achieving credibility.