Non-negative Matrix Factorization And Its Application In Personalized Recommending System

Matrix factorization has been previously shown to be a very useful tool for data min-ing and data analysis. Recently, works in machine learning has focused on matrix fac-torizations that directly target some of the special features of statistical data analysis. Inparticular, nonnegative matrix factorization (NMF) focuses on the analysis of data matri-ces whose elements are nonnegative, which allow to extraction feature easily. The scope ofresearch on NMF has grown rapidly in many applications, such as environment, patternrecognition, multimedia, text mining, and DNA gene expressions. In numerical linear alge-bra, negative matrix factorization (NMF) is a popular dimensionality reduction techniquethat allowed many huge and complex problems decompose into small and easy problemsto solve; in machine learning and pattern recognition area, NMF applied to high dimen-sional data where each element is nonnegative and it provides a low rank approximationformed by factors whose elements are themselves nonnegative, it reduced the storage andcomputation resource.E-Commerce companies compete each other nowadays,the E-Commerce grows moreand more quickly, the trend of this industry also changes very fast. Since the E-Commercecompanies can provide a great variety of products, the customers not only can buy some-thing have obvious common features, but also can buy something seems nothing interre-lated in objective. As a manager of an E-Commerce company, he need to get knowledgeabout customers’ preference in order to establish beteer adequate personalized services tosatisfy the customers.We interpret the factorization in a new way and apply it to personal Recommenda-tion as a clustering method in electronic commerce. As the NMF method can extract betterimplicit feature in customer data, we can improve the accurate of the customers’ cluster-ing, and also recommend the right product to every customer in a short time.