Abstract: [Purpose/Significance] The purpose of this study is to improve the information service level of online health community and meet the deep information needs of patients and users in different scenarios. [Design/Methodology] Integrated the scenario elements of information recommendation service, based on the analysis of the information needs of users in the online health community, the scenario division criteria were determined and then the recommendation scenarios were divided, so as to identify different scenario element labels to build a user portrait conceptual model. By RFM model, the user data of the online health communities were refined and operated, and the user portrait model of the chronic disease online health community was constructed by means of the method of formal concept analysis. In addition, "Sweet Homeland" was used as the data source to realize the user portrait construction and label clustering of diabetes online health community. [Findings/Conclusion] Through in-depth mining and clustering, this paper finds four types of user groups and their health information needs namely "wait-and-see","propagandistic","prolific" and "experience". [Originality/Value] Under the guidance by the information recommendation service model, this paper focuses on the overall process of user portrait construction, which is helpful to optimize the service level of online health community and improve user satisfaction.