The use of rough sets theory in multi-criteria recommender systems

Accessing the information has become quite easier with the development of technology and the discovery of the Internet but the opportunity to get billions of data in seconds brought the problem called Information Overload with it. In order to overcome this problem, recommender systems that predicts the necessary information from tons of data has been proposed. Within the dozens of different approaches for recommender systems, collaborative filtering is one of the most commonly used systems for making predictions. In collaborative filtering recommender systems, the users put forth their characteristics by giving rating to the products they know and the system produces prediction to the products they do not have any idea in return. In traditional collaborative filtering, the products are considered as a whole, but sometimes approaching to a product from multiple aspects makes the system more successful. In this context, multi-criteria recommender systems are developed. In this PhD dissertation, in-depth study of multi-criteria recommender systems is performed and an aggregation function-based multi-criteria collaborative filtering using Rough Sets Theory is proposed as a novel approach. The traditional collaborative filtering systems have problems such as scalability and sparsity. Since the multi-criteria collaborative filtering systems are considered as the extended version of the traditional single-rating collaborative filtering systems, these problems pose a much grater problem. Additionally, revealing the relation between criteria is yet another optimization problem. Hence, increasing the accuracy in prediction is a challenge. In this dissertation, a number of approaches have been made to improve the predictive accuracy of multi-criteria recommender systems. A novel approach which uses Rough Sets Theory has been proposed to overcome the challenges of these systems.