SVD based Privacy Preserving Recommendation Model using Optimized Hybrid Item-based Collaborative Filtering

Abstract

Recommender system is one of the most decision support system which solves the problem of filtering required information from huge collected information as per the user's interest, choice or item's property. Collaborative based filtering recommender system is one of best filtering approaches which is very effective in wide range applications. Item Based Collaborative Filtering (IBCF) approach solves scalability and data sparsity issue with better accuracy. In this paper, we have developed optimized hybrid item based collaborative filtering recommendation model using binary rating matrix along with Jaccard similarity and used basic Singular Value Decomposition method (SVD) for privacy preserving. In this model, binary rating matrix is first examined and relationships among various items are identified by optimizing nearest neighbors as parameter. Then we use these relationships that help with recommendations for the user by achieving high privacy and better accuracy of the model.