期刊名称：International Journal of Innovative Research in Science, Engineering and Technology
印刷版ISSN：2347-6710
电子版ISSN：2319-8753
出版年度：2017
卷号：6
期号：8
页码：17389
DOI：10.15680/IJIRSET.2017.0608259
出版社：S&S Publications
摘要：Collaborative filtering is an efficient recommendation approach during which the preference of a user onan item is predicted supported the preferences of different users with similar interests. A big challenge in exploitationcollaborative filtering ways is the information sparsity drawback which frequently arises because every user generallyonly rates only a few things and therefore the rating matrix is extraordinarily thin. In this paper, we address thisdrawback by considering multiple collaborative filtering tasks in several domains at the same time and exploiting therelationships between domains. We refer to it as a collaborative filtering (CF) drawback. To solve the CF drawback, wetend to propose a probabilistic framework that uses probabilistic matrix factorization to model the rating problem ineach domain and permits the information to be adaptively transferred across totally different domains by automaticallylearning the correlation between domains. The proposed framework of DsRec includes three components: a matrixfactorization model for the determined rating reconstruction, a bi-clustering model for the user-item subgroup analysis,and two regularization terms to connect the above 2 components into a unified formulation. In existing we tend to hadtaken movie information and analysis subgroup analysis in our proposed system we had taken, multiple product itemsand analysis subgroup analysis.
