Research On The Application Of Non-negative Matrix Factorization And AP Clustering Methods In Personalized Recommendation Systems

With the rapid progress of Web2.0 era,we gradually from the era of poor information resources into the era of the proliferation of information resources.The ever-changing speed blinds the user's eyes,looking for useful information to consume a lot of energy users,the same,a lot of information or goods in trouble.When the user can not get from the massive network resources to obtain the information they need,personalized recommendation system to the role of sales staff in the world before the timely and accurate grasp of user needs,and truly information,goods and users of the three Between the seamless docking.From the perspective of the development process of the recommendation system,the following problems are prevalent:the system can not make recommendations for new members,and the problem of sparseness of user's evaluation data of goods is poor as the number of users and projects increases.The Based on the above problems,this paper analyzes the research status and application value of nonnegative matrix decomposition(NMF)and clustering technology,improves the nonnegative matrix decomposition algorithm,introduces the near-neighbor propagation clustering algorithm(APC),proposes the modified NMF Algorithm and proximity propagation clustering(APC),which aims to solve the sparseness and expansibility of personalized recommendation system data.The main work of this paper is as follows:1.On the basis of analyzing the development status of personalized recommendation system,we try to give the research focus,difficulty and hotspot of recommendation system.2.(APC)is studied in detail.Based on this,the recommendation algorithm of APC algorithm and collaborative filtering is studied and compared with the popular clustering algorithm K-means.It is shown that AP clustering Methods in dealing with high-dimensional,sparsely populated data.(NMF)is proposed in this paper.A modified NMF method is proposed,which combines the technique with cooperative filtering.The iterative method makes the decomposition matrix fully fit the original matrix and effectively solves the problem of data sparseness.3.On the basis of the above research,this paper proposes a recommendation algorithm combining APC with NMF.The original score matrix is decomposed by NMF to produce user factor and project factor matrix.The APC technique is used to cluster the two matrices respectively.Classify different users or project types,and finally through the collaborative filtering algorithm to predict the score,and then produce recommendations.4.According to the combination recommendation algorithm proposed in this paper,combined with the evaluation index of the recommendation system,the off-line data and the traditional algorithm are used to test the comparison,which proves that the proposed algorithm is superior to the traditional algorithm in the recommended speed and accuracy.