[Purpose/significance] According to the social identity theory,the behaviors and attitudes of the netizens in using micro-blog are closely related to their perception of their membership. Building users’ portraits of all kinds of groups in micro-blog can directly show the typical characters of the group members which help to analyze users’ behaviors and attitudes in various kinds of groups,and are important in the network public opinion management,advertising,personalized service,and other marketing research. [Method/process] Firstly,using social identity theory,the paper finds out users’ interested micro-blog topics through thematic model text mining. Then,user similarity is calculated using the space vector of users’ preferences in topics probability distribution,thus realizing the classification of user groups. Finally,the paper extracts the main feature attributes for different groups.[Result/conclusion]Five typical groups are found in micro-blog and members of certain groups are easily affected by the outside groups and further make adjustments to reach new positive identity. The typical user portraits of different micro-blog groups can be obtained. According to the main characteristic attributes of groups,the paper analyzes the user behaviors,attitudes,and gives the corresponding advices of network public opinion management,personalized service and marketing strategy.