Controllable Uncertain Information Diffusion Influenced By Individual Opinion

Analyzing and controlling the diffusion process in social network is important in the context of viral marketing and political propaganda. Recently, many studies focus on uncertain information diffusion. The uncertainty of diffusion process means that the diffusion probability is unpredicted because of some complex factors. For instance, individual opinion is an important factor that can cause uncertainty of diffusion probability. Individuals’ opinions widely exist in social network and significantly influence the diffusion process. However, previous studies have not shown how to control the uncertain diffusion process influenced by individual opinion. In this paper, we present strategy to control this diffusion process based on the approximate estimation of the uncertain factors.First, we formulate a model in which the diffusion process is influenced by individual opinion, and we discuss the properties of the diffusion model by theoretical analyses and various simulations. Results show that, if the range of opinions that can be accepted by the social individuals is larger, the final scale of diffusion will also be larger. Moreover, if the initial distribution of opinions is convergent, the final scale is larger.Second, we analyze the properties of the diffusion process from another angle. In detail, we discuss the evolution of opinions in diffusion process. It is shown in results that network structure can only influence the degree of opinion evolution. However, the range of opinions that can be accepted will not only influence the degree but also the tendency of opinion evolution.Finally, we present an opinion influenced maximization problem based on the voting in real world. In detail, we model a voting mechanism in which individual can choose two candidate or abstention based on his opinion. Then, we present strategies to select suitable initiators and their opinions in diffusion process so that the advantage of one candidate will be maximized. The results show that the estimation value of initiator’s most suitable opinion is close to the simulated value. Besides, traditional influence maximization algorithm is not applicable in this problem, and our heuristic algorithm can achieve the desired results.