Neighborhood-based Collaborative Filtering Using Grey Relational Analysis

摘要

The popular neighborhood methods in collaborative filtering are usually used to recommend items that users with similar preferences have liked in the past. The similarity indicates that there exist a certain degree of relationship between one user and other users. Since the grey relational analysis (GRA) is an effective technique that can measure the degree of relationships among patterns for multi-criteria decision making (MCDM), this motivates us to use this technique to design the similarity measure for neighborhood methods. The proposed similarity of one user to another user is thus dependent on the strength of the relationship between the former and the latter. In contrast to traditional similarity measures for neighborhood methods in collaborative filtering, the proposed similarity is not symmetric for any two users. The applicability of the proposed single-criterion and multi-criteria similarity-based methods to the recommendation of initiators on a group-buying website is examined. Experimental results have demonstrated that the generalization ability of the multi-criteria neighborhood method using the proposed similarity performs well in comparison to that using other similarity measures.