Model of warning information diffusion on online social networks based on population dynamics

Abstract

With the development of Internet technology and mobile terminals, online social networks have played an increasingly important role in warning information diffusion. In this paper, a population dynamics model is introduced to describe the warning information diffusion process on online social networks. This model takes information attraction, network capacity, warning timeliness, and the interaction between different information into consideration. It could be used to describe and predict the diffusion process of a piece or pieces of warning information without the necessity to capture the detailed structure of social networks. A series of simulation is conducted to reveal the mechanism of the parameters influence on the diffusion process. This model is verified by real-world cases of warning micro-blogs on Sina Weibo about Super Typhoon Licma and results show that this model could well describe the diffusion process of warning information.

Introduction

With the development of network technology and the popularization of mobile terminals, online social networks such as Facebook, Twitter and Sina Weibo have become the mainstream information dissemination medium in the current era.

According to Facebook’s fourth quarter 2018 financial report, the number of its daily active users reached almost 1.5 billion [1]. Data center of Sina Weibo also claimed that their daily users have reached 200 million in their yearly report of 2018 [2]. A huge number of users brings not only a more complex network environment but also faster diffusion speed and wider diffusion range of information. As is well known, information is mixed by truth and rumors. In order to stop rumors and promote important notice like early warnings of natural hazards, it becomes more and more urgent to understand the information diffusion process on online social networks.

By now many works have been done to reveal the diffusion mechanism of the information on online social network. Some of them focus on the explanatory aspect of the diffusion process. For example, Hui et al. put forward a diffusion model of actionable information to describe the message spread process on Twitter [3]. Gomez proposed that the infection times could be one of the important factors that influence the infecting process and developed an algorithm called NETINF based on it [4], [5]. Haralabopoulos et al. looked into the information connection in various online social networks and provided a measurement of shared information lifetime in multiple networks [6]. Compared with diffusion mechanism of the information, current research paid more attention to predictive models for their larger scope of application, like Yang and Leskovec’s Linear Influence Model [7], Matsubara’s SpikeM model [8], Jiang’s evolutionary game theoretic framework [9], [10], Sutton’s a negative binomial regression model [11], and some other models [12], [13], [14].

One of the most popular models among them is the epidemic model, such as SIR and SIS models. Referring to the dynamics of infectious diseases, the SIR and SIS models divide user nodes into three categories: susceptible(S), infected(I) and recovered(R), and the information is analogized to the source of infection to construct a dynamic model [15]. For simple situation, the information diffusion process could be calculated by mathematical methods. While for more complicated scenarios, the network structure should be learned to conduct a more elaborate and accurate simulation. Besides, in some research it is found that different information is interacting with each other during diffusion [16], while epidemic models cannot deal with the situations when multiple information is concerned simultaneously.

In this paper we propose a new information diffusion model based on the population dynamics, which could roughly predict the diffusion process of multiple information without specific network structure. Population dynamics model is a mathematical model used for describing the relationship between populations, including competing and cooperating relations with each other, and predicting the change of population size over time. In recent years it has been applied in medicine, demography, sociology and other areas. Some researchers applied the logistic function of population dynamics into the spreading of ideas, knowledge and innovations [17], [18]. Besides, Wang et al. promoted Diffusive Logistic Model based on population dynamics to model the impact of user influence within the same group over online social networks, and empirical experiments showed that the DL model could effectively predict the density of influenced users for a given distance and a given time for both distance metrics [19].

By comparison, epidemic models are particular cases of population dynamics models. They both use partial differential equations to describe the diffusion process of information, idea and innovation, but epidemic models emphasize network nodes in single message’s diffusion while population dynamics models pay attention to the overall trend of information spreading [18]. Thus epidemic models are more suitable for single message with clear diffusion network, and population dynamics could handle multiple information problems and give better prediction in implicit network.

This paper mainly focuses on the diffusion of warning information. Compared with normal information, warning information is different due to its characteristics of time-efficient, authoritative and significant effect on reducing losses during emergencies. By learning the competitive and cooperative relationship between warning information and other information, we could promote the diffusion of warning information when needed, and eventually reduce losses when emergency happens. Considering the advantages of population dynamics when describing multiple objects, we establish the warning information diffusion model based on population dynamics models.