The invention provides a user big data based user group analysis method and system. The method comprises: constructing a relationship network of users and user tags; representing the users by nodes in the relationship network, and if two nodes have the same tags, establishing a connection relationship between the two nodes, wherein the weight of a connection line of the two nodes is the number of the same tags between the two nodes; traversing the relationship network, combining two nodes corresponding to a connection line with the maximum weight in the network into a cluster, and taking the cluster as a new node; calculating weights of the cluster and connection lines of other nodes in the network; and if the updated relationship network still can perform node combination, continuing to obtain new clusters until the number of clusters in the updated relationship network reaches a preset value, wherein nodes contained in the clusters in the updated relationship network are discovered similar user groups. According to the technical scheme, the analysis on the user groups is more intuitive.