In-Depth Analysis of Similarity Knowledge and Metric Contributionsto Recommender Performance

May 2004

Proceedings of the Seventeenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2004)

Volume

Proceedings of the Seventeenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2004)

All Papers

Initial successes in the area of recommender systemshave led to considerable early optimism. However asa research community, we are still in the early days ofour understanding of recommender systems. Evaluation metrics continue to be refined but we still needto account for the relative contributions of the variousknowledge elements that play a part in the recommendation process. In this paper, we make a fine-grainedanalysis of a successful approach in the area of casebased recommendation, providing an ablation studyof similarity knowledge and similarity metric contributions to improved system performance. We gaugethe strengths and weaknesses of knowledge componentsand discuss future work as well as implications for research in the area.