In the existing user modeling methods for collaborative tagging systems,a user is regarded as a tag- vector and it is assumed to be interested in every tag in the tag- vector.Moreover,only the matching degree of a tag with another tag is calculated,while the effects of tags as a whole on the user’ s preference are ignored.In order to solve these problems,this paper proposes a recommendation algorithm based on the tag- group effect,namely,TGER.This algorithm utilizes the user ratings on resources to select the tag- groups which have significant effects on the user’ s preference,and adopts the high- dimension tag- group first matching method to calculate the user- resource relevance.Experimental results on the MovieLens data set show that TGER can significantly improve the recommendation qua-lity.