[Purpose/significance] Online health community has been developing rapidly, but it is difficult to reveal the inherent mechanism between network node attribute and user dynamic interaction mode, with the lack of research on node attribute-based dynamic networks.[Method/process] This paper constructed 155 the research model for users temporal interactive mode of online health community by applying NATERGM（Node Attribute-based Temporal Exponential Random Graph Model）, taking the online health community as research object, and combining with the feature of the user node properties that including users emotional tendency, the post text and the user community level.Then, data from Baidu Diabetes community from October 2016 to February 2018, which involves 2301 users, 6045 posts and 9490 replies, was used for empirical research.[Results/conclusion] Results indicate that the validity of NATERGM in analyzing temporal interactive mode, and node attributes have significant effects on reciprocity temporal mode, k-star temporal mode, transitivity temporal mode and cyclicity temporal mode.
Consequently, suggestions for the Circular/DIN development of online healthy community construction are put forward.