Investigation for Designing of Context-Aware Recommendation System Using SVM

Abstract Previously, we have proposed two recommendation systems, the Context-aware Information Filtering (C-IF) and Context-aware Collaborative Filtering (C-CF), both of which are contextaware recommendation methods. We have also shown their eectiveness through of experiments using a restaurant recommendation system based on these methods. However, we have not discussed how to eectively and suitably develop a practical contextaware recommendation system. In this study, we analyze the following: a) the appropriateness of adopting a Support Vector Machine (SVM) to a context-aware recommendation method; b) advantages and disadvantages of C-IF and C-CF in various user situations; and c) optimization of parameters of feature vectors for target contents and user contexts. As a consequence, we veried that it is appropriate to adopt an SVM to a recommendation method since the SVM has high generalization performance. We discovered advantages of both the C-IF and the C-CF in dierent recomm...

• Type: Conference
• Year: 2007
• Where: IMECS
• Authors: Kenta Oku, Shinsuke Nakajima, Jun Miyazaki, Shunsuke Uemura
• Keywords: Contextaware Recommendation, IMECS 2007, Recommendation Method, Recommendation System, Theoretical Computer Science