Learning User Similarity and Rating Style for Collaborative Recommendation

Abstract

Information filtering is an area getting more important as we have long been flooded with too much information. Product brokering in e-commerce is a typical example and systems which can recommend products to their users in a personalized manner have been studied rigoriously in recent years. Collaborative filtering is one of the commonly used approaches where careful choices of the user similarity measure and the rating style representation are required, and yet there is no guarantee for their optimality. In this paper, we propose the use of machine learning techniques to learn the user similarity as well as the rating style. A criterion function measuring the prediction errors is used and several problem formulations are proposed together with their learning algorithms. We have evaluated our proposed methods using the EachMovie dataset and succeeded in obtaining significant improvement in recommendation accuracy when compared with the standard correlation method.