| Currently, behavior analysis based on web social behavior is becoming the focal point of academic research and commercial operation. It has been applied in many fields,such as personalized recommendation, online marketing, public opinion analysis and forecasting.Existing studies generally indicate that online community possesses the features of small world and scale-free. And whether user interaction is assortative or not is related to the type of the network community. However, researches on user behavior and online community are concentrated in large-scale comprehensive social network site and small-scale BBS system. There are less researches focus on general characteristics analysis of regional network community user behavior. Unlike large-scale comprehensive social network, regional network community provides a platform for users who have the similar range of geographical and cultural backgrounds. It is similar with the small-scale social network, such as university forum and professional BBS. But regional network community has relative large number of users. And the user population social attributes are more diverse. Thus, researches of the regional network community user behavior will contribute to summarize the general characteristics of different type of social network platform.This paper selected Chengdu local online communities as the research objects.Through questionnaire and the analysis of platforms, two representative regional network communities, "XCAR Chuan sub-forum" and "Just Chengdu", were selected as study objects. This paper utilized statistical analysis, social network analysis and assessment analysis to study user behavior characteristics and user influence. The specific studies are as follows:Firstly, the overall development statistics of two network communities were carried out. And through analysis and comparison of two online communities, the user active time distributions were obtained, which provide the support for online public opinion monitoring.Secondly, the user interaction social network was built by analyzing the information dissemination pattern of online network community. Through comparative analysis of the distribution characteristics of the user degree, the user interactionfeatures of regional network community were found. By analysis of user interaction propensity from out-degree and in-degree, it found that the regional network community is disassortative. It supports the emotional reasons for the development of regional network community.Finally, based on the 5W model of information dissemination, the formation of user influence was analyzed. From the perspective of user reply and network co-link, it created the model to assess the user influence and classify the high-impact users by using PageRank algorithm. This method improves the existing study of user influence evaluation. |
