Abstract: Now it's much more common that people participate in online public opinion. But because of the imperfect network public opinion management system, the network public opinion events can make serious and negative effects easily. So the importance of studying network public opinion is increasingly significant. The study of the complex online public opinion based on evolutionary game theory can reflect the actual situation of the online public opinion formation and evolution. It helps to provide some scientific basis for guiding the users' behaviors and controlling the evolution of the online public opinion. This article introduces two factors into the replication dynamic model of online public opinion which can reflect the complex interaction behaviors between the internet users:the preferences of creating a new game connect and the lifetime of maintaining game connect. An evolutionary game model of network public opinion on this basis is built. According to evolutionary game equilibrium solution, how the complex interaction factors and the initial benefit influence the evolution of network public opinion is analyzed and explained. Besides, the corresponding suggestions about the network public opinion management for different situations are put forward. The study shows that when the interactive connections reach steady state, the gain matrix of the network public opinion will change. The elements in the original benefit matrix times its corresponding active connection proportion of the overall population under the different game connection type then the new benefit matrix is composed of the news elements. The complex interactions among the internet users are quantitatively explained. The study plays an important role in many aspects such as giving advices about online public opinion crisis management, reducing the potential hazards of online public opinion.

Key words: online public opinion, evolutionary game, interaction behavior