A Comparison Study of Different Privacy Preserving Techniques in Collaborative Filtering Based Recommender System

Abstract

Recommender System is a productive tool which has been used in various fields for personalizing the applications so that we get recommended items such as e-commerce sites, shopping sites like Amazon, Netflix, research journals, mobile-based platforms and TV-based platforms. Collaborative Filtering (CF) is the most popular widely used tool on the Internet for recommender systems to search for new items that fit users’ interest by finding similar users’ opinion expressed on other similar items. However, there are many disadvantages for CF algorithm, which are data privacy risk and data sparsity. In this paper, we compare and analyze different privacy preserving collaborating filtering techniques that have greatly captured the attention of researchers i.e. Matrix Factorization (MF) and Singular Value Decomposition (SVD) with finding accuracy. Compared with SVD and MF, WBSVD achieves better recommendation quality and is more accurate from experimental results.