Abstract: Under the Internet environment, consumers gradually shift from traditional individual decision-making to group sharing and collaboration, and the online community has become an important source of information for their observational learning. Considering its gradual and phased characteristics, this paper constructs the theoretical framework of network community learning from three levels: “group recognition”, “member participation” and “viewpoint distribution”. Secondly, web crawlers are used to capture interactive data of online communities, and machine learning algorithms are used to analyze and process unstructured data. Finally, the dynamic panel econometric model is established based on the system moment estimation method. It is found that “group recognition”, “member participation” and “viewpoint distribution” all have positive effects on the community learning effect in the network community, while “member participation” and “viewpoint distribution” have progressive positive moderating effects on the above effects. This paper attempts to open the “black box” of community learning mechanism and provide micro theoretical support for the burgeoning community economic development.

Key words: network community, community learning, learning effect, text mining, community economy