Item-Based Recommender Systems Applying Social-Economic Indicators

Abstract

Despite the development of online items in the market, Item-based Recommender Systems play an essential role to assist consumers to access their targets rapidly. Research activities show that users (consumers) in various societies have different purchasing behaviors. In the latest studies, the similarity of users was calculated based on different indications such as Cultural Indicators. Nonetheless, they neglect social-economic indicators expected to be more factors that are efficient. In this research, a framework was designed for recommender systems called SEIRS that include several algorithms based on social-economic indicators. In fact, the similarity of items was defined in terms of the similarity between users’ evaluation based on social-economic indicators. The presented method categorizes items and recommends them to users so that they match with users’ favorites more precisely. The results were compared not only with the standard item-based system but also with different mixtures of social-based and economical-based indicators. In addition, the presented idea was used to solve the cold start problem that occurs when a newcomer customer with no information on his previous purchases and evaluations becomes available. Furthermore, it is possible to present suggestions that are more precise by using the SVD algorithm to solve the data sparsity. The present research applied three datasets: first, internet purchase by individuals (isoc_ibuy) dataset, second the related data on social-economic indicators from OECD, and third the actual BookCrossing dataset. The experimental results represent that in comparison with the standard recommender system, by applying social indicators the proposed recommender algorithm improved precision by 17.19%, and economic indicators improved precision by 17.60%, respectively. Applying both social and economic indicators provides 20.91% more accurate recommendations.