A Survey of Collaborative Filtering-based Systems for Online Recommendation

Internet services operate on a vastly larger scale and permit virtual interactions. The Internet and Web has created vast new opportunities, providing an infrastructure that enables buyers and sellers to find each other online. Companies can now offer many products, services and information easily and with lower costs. It becomes more and more difficult for customers to find quickly what they are looking for. Nevertheless, recommendation systems are playing a major role. Collaborative filtering (CF), or recommender system based-CF, has appeared as one methodology designed to perform such a recommendation task. These systems allow people to use expressed preferences of thousands of other people in order to find the product they desire based on the level of similarity between tastes. The concept has appeared from convergent research on search browsers, intelligent agents and data mining, and it allows to avoid the difficult question of "why" consumers prefer this or that product or brand. Early studies of electronic markets tools and recommender systems took a simplistic view of consumers as economic agents whose behavior was guided by the search for the lowest cost transactions. Moreover, most studies take into account only technical aspects of these systems like algorithms' development and computational problems. No study had been interested in recommendation's efficiency of collaborative filtering-based systems. This article explores the current state of research in recommender systems-based collaborative filtering, and proposes an experiment to find if such electronic recommendations are better than human recommendation.