The invention discloses a community mining method based on local similarity. According to the community mining method based on local similarity, local areas in close relation are found out according to the closeness of the relation of a network structure, all the areas are taken as prototypes of communities and labels are assigned to the areas, and the areas are reexpanded and adjusted according to the vote principle by means of a label spreading method so that the communities can be formed. According to the community mining method based on local similarity, no manual parameter is needed, and a community structure is obtained directly according to the structural property of the closeness of the relation between the nodes of the network.