Abstract: The advent of big data age makes the influence of public opinion leaders and ordinary users significantly different. On the basis of " one-to-many” information dissemination mode between public opinion leaders and ordinary users, this paper analyzes the changes of user behavior with time at the macro level, and at the micro level, it analyzes the characters of ordinary users with the help of community detection algorithm. To carry out empirical analysis, two cases as test samples are selected, and it is found that the " many-to-many” interaction mode exists under the influence of the " one-to-many” mode. The results show that the " many-to-many” mode has a shorter interaction period, and the Pareto law can be applied to communities, which are formed through interaction.