Improving accuracy of privacy-preserving collaborative filtering methods by rough sets theory

Recommender systems have become popular with increasing use of Internet. Collaborative filtering methods which use recommender systems have been used in order to for selecting product over online platforms. There are some challenges such as privacy, accuracy, online performance, coverage, sparsity and scalability of these methods. Rough sets theory has been used in order to overcome these challenges. Rough sets theory has been used with the intent of improving accuracy, expansion of coverage and increasing online performance of the collaborative filtering methods. Privacy is one of the important issues of the collaborative filtering methods. After customers log in the system, this issue begins. Privacy-preserving collaborative filtering methods have been used to cope with this issue. However, these methods decrease accuracy of the recommender system. The purpose of this thesis is overcoming the issue of preserving of privacy. ROUSTIDA algorithm which improved using indiscernibility relation in the rough sets theory has been used to cope with this issue. This approach has been tested with three different algorithms, namely memory-based privacy-preserving, model-based privacy-preserving and hybrid privacypreserving. In result of experiments, accuracy and sparsity issues have been relieved in the privacy-preserving collaborative filtering methods.