An Ontological Sub-Matrix Factorization based Approach for Cold-Start Issue in Recommender Systems

With the rapidly growing usage of e-commerce applications, it is becoming more and more tedious for the vendors to perform accurate and relevant recommendations to the users visiting their websites, especially first time. This is a type of cold start problem in the recommender systems. In this paper, an ontological sub-matrix factorization based approach is suggested for recommending items to a new user. The main contribution of the paper is that, for recommending any items to a new user, no personal information regarding user is captured or extracted, thereby respecting the privacy of the user. The proposed approach when tested has shown an accuracy of 98 percent in terms of recall value.