A New Similarity Measure Based on Preference Sequences for Collaborative Filtering
Collaborative filtering is one of the most popular techniques in recommender systems, and the key point is to find similar users and items. There are already some similarity measures, such as vector cosine similarity and Pearson’s correlation coefficient, and so on. However, in some cases, what recommender systems get are not the ratings, but preference sequences of users on a series of items. For this type of data, those traditional similarity measures may fail to meet the practical application requirements. In this paper, a similarity measure based on inversion is proposed for preference sequences naturally. Based on the Inversion similarity measure, some structural information of user preference sequences is analyzed. By merging average precision and weighted inversion into similarity computation, a new similarity measure based on preference sequences is proposed for collaborative filtering. Experimental results show that the proposed similarity measure based on preference sequences outperforms the common similarity measures on the datasets with continuous real numbers.
KeywordsRecommender System Collaborative Filtering Similarity Measure Preference Sequences Weighted Inversion Average Precision
Unable to display preview. Download preview PDF.
- 3.Chowdhury, G. Introduction to modern information retrieval. Facet Publishing (2010)Google Scholar
- 4.Cormen, T., Leiserson, C., Rivest, R., Stein, C.: Introduction to algorithms, 3rd edn. MIT Press and McGraw-Hill (2009)Google Scholar
- 9.Su, X., Greiner, R., Khoshgoftaar, T., Zhu, X.: Hybrid collaborative filtering algorithms using a mixture of experts. In: Proceedings of the IEEE/WIC/ACM International Conference on Web Intelligence, pp. 645–649. IEEE Computer Society (2007)Google Scholar
- 10.Su, X., Khoshgoftaar, T.: A survey of collaborative filtering techniques. In: Advances in Artificial Intelligence 2009, p. 4 (2009)Google Scholar