Finding Similar Users over Multiple Attributes on the Basis of Intuitionistic Fuzzy Set

Finding similar users is vital for many applications, such as collaborative filtering and recommendation systems. Unlike existing work that can only discover similar users according to one attribute, we propose a method based on intuitionistic fuzzy set for finding similar users according to multiple attributes: interest, behavior, and personal information. For a single attribute, such as user interest, the interest of two users may be partly similar, different in another aspect, and fuzzy in the remaining part. The three parts correspond exactly to the three concepts of the intuitionistic fuzzy numbers: membership, non-membership, and uncertainty degrees. And the integrated operator of the intuitionistic fuzzy set is just used for considering all attributes comprehensively. Based on this idea, we firstly introduce the method for determining the intuitionistic fuzzy numbers of each attribute. And then we use intuitionistic fuzzy hybrid average operators to integrate them to quantify the user similarity level. And we can further identify the similar users in terms of multiple attributes based on the level. The experiments based on Twitter data show our method outperforms the baselines.