Abstract:  With the increase of resource-sharing web sites such as delicious, Flickr, last.fm, allow users tagging the contents that they interested in. It is important to discovering social interests shared by groups of users because it helps to search for people with common interests and encourages people to generate and share more resources. In those Folksonomy, Tags can be used to describe the content of resources, and also can represent user’s interests. We use tags to represent user’s interest，and three notions: Degree of annotating consistency (DAC), Degree of resource sharing between friends (DRSF) and Degree of interest similarity between friends (DISF) are defined, DAC are used to construct user’s interest model. According to the statistical experiment, we find that the result of user’s DISF is higher than the result of their DRSF in folksonomy. So we consider adding DISF into the calculation of users’ interest similarity. We use this new user’s interest similarity calculating method in community detection algorithm (SCAN), the experiment results show that the proposed method is effective.