A Goal-Driven Attribute Selection Method for Recommendation Systems

Abstract

This paper reports a requirements engineering approach to attribute selection for enhancing the results of a recommendation system. A recommendation system suffers the sparsity problem and the cold start problem with collaborative filtering which are caused by the lack of data. Our method is to introduce more timely information of user preferences to enhance the recommendation results that meet the current needs of a user. The proposed method uses a goal-driven approach with the support of the Analytic Hierarchy Process in attribute selection. The experiments show that this method derives promising results.