Exploring Dissimilarity in Collaborative Recommendation Systems

• Lucas Miranda UFMG
• Fernando Mourão UFMG
• Wagner Meira Jr. UFMG

Abstract

The huge amount of options available in various commercial applications became Recommender Systems (RS) crucial tools to assist users in their choices. Despite recent advances in RS, there is still room for more effective techniques which are applicable to a larger number of domains. Most problems arise from the simplified model recurrently used. In this paper, we propose a richer user modeling which allows to extrapolate the usual similarity analysis. Furthermore, we propose a technique that, by exploiting an information type defined as dissimilarity, provides significant improvements over traditional techniques based on collaborative systems, as well as reduces the analysis cost required by such techniques.

References

Adomavicius, G. and Tuzhilin, A. (2005). Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE TKDE, 17(6):734–749.

Breese, J., Heckerman, D., and Kadie, C. (1992). Empirical analysis of predictive algorithms for collaborative filtering. Learning, 9:309–347.

Burke, R. (2002). Hybrid recommender systems: Survey and experiments. User Modeling and User-Adapted Interaction, 12(4):331–370.

Han, J., Pei, J., and Yin, Y. (2000). Mining frequent patterns without candidate generation. ACM SIGMOD Record, 29(2):1–12.

Hu, Y., Koren, Y., and Volinsky, C. (2008). Collaborative filtering for implicit feedback datasets. In 8th IEEE ICDM, pages 263–272.

Kagie, M., Van Wezel, M., and Groenen, P. An Empirical Comparison of Dissimilarity Measures for Recommender Systems.

Koren, Y. (2009). Collaborative filtering with temporal dynamics. In Proc. of the 15th SIGKDD, pages 447–456. ACM New York, NY, USA.

Park, Y. and Tuzhilin, A. (2008). The long tail of recommender systems and how to leverage it. In Proc. of the 2008 ACM RecSys, pages 11–18. ACM.