Continous Opinion Dynamics Evolution On Social Networks And Its Application In Online Prediction

In recent years, the research methods in natural sciences which are used to study complex social group behaviors and phenomena have aroused widespread interests and concerns. Opinion dynamics utilizes the agent-based modeling and simulation to study the transition process from disorder to order of individuals’ opinions in a social group. An opinion is the individual idea, choice, preference about some problem or issue. Opinion dynamics considers that the formation and evolution of the individual opinion are influenced by his opinion and the opinions of others around him. It tries to establish the model of how opinions influence each other, to analyze and interpret the spread, evolution, clusters emergence of opinions. In depth study on opinion dynamics has important theoretical and practical significance to deepen people to understand the mechanism of formation and evolution, design better decision-making and discussion process for large-scale crowds, guide and control the spread of public opinion and the formation of consensus, and so on. This paper studies issues about theory and application of continuous opinion dynamics evolution, including the following four aspects:1. The research on opinion formation of free speech on the directed scale-free social network. The interactions of individuals in a social group form a directed scale-free network. The individuals express their opinions one by one with random order(RO) or probability order(PO), others paying attentions to the speaking individual, receive provider’s opinion, update their opinions and then express their new opinions in their turns. A continuous opinion dynamical model on a directed social network based on the hypothesis of ”bounded confidence” is proposed to study how the speech order and the topology of the social network affect the opinion formation and evolution. The results show that: 1) with the same individual repeats its opinion more, others who pay their attentions to him and include his opinion in their confidence levels at initial time, will continue approaching his opinion; 2) the model for PO forms fewer opinion clusters, larger maximum cluster, smaller standard deviation, and needs less waiting time to reach a middle level of consensus than RO; 3) as the parameter of scale-free degree distribution decreases or the confidence level increases, the results often get better for both speech orders; 4) the differences between PO and RO get smaller as the size of network decreases.2. The research on opinion evolution of round speech. Most dynamical models with continuous opinion lack the considerations of trust between individuals and different weights of opinions. The extended Hegselmann-Krause(HK) model of opinion updating with weights is proposed by introducing the trust between individuals and the similarity between opinions, and extending the hypothesis of bounded confidence to bounded influence. The model is used to study the formation, evolution of opinions in a social group and consensus-building process under the influence of a few of narrow-minded and authoritative individuals. Simulation results show that: decreases the difference between the initial opinions of two types of individuals and the midpoint of distribution range(0.5), or increases the influence thresholds of narrow-minded individuals and the trust degree of authoritative individuals would form larger and fewer opinion clusters. Compared the variations of round speech with free speech, simulation results show that under the same initial conditions, the variation of free speech taking the speech sequence rule of authoritative individuals first and then individuals with the maximum of opinion cumulative change first has less opinion clusters, and 15 rounds of speech is the threshold to determine the change of opinion clusters.3. The study of coevolution of continuous opinions and directed adaptive network. The Hegselmann-Krause(HK) model is extended to investigate the coevolution of continuous opinions and directed adaptive Erd?s-Rényi network. Directed links can be broken with a probability if the difference of two opinions exceeds a certain confidence level e, but new links can form randomly between two individuals without a directed link. Simulation results reveal that on static networks, final opinions are influenced by percolation properties of networks; but on directed adaptive networks, it is basically determined by the rewiring probability. The increasing rewired probability increases the average degree of network, enhances communications between agents, which lead that the results of adaptive networks are getting better than static networks.4. The study of number prediction for online rating based on continuous opinion dynamics. On the process of online rating, the individual opinion is influenced by his initial opinion and the group’s opinions. The final opinion will determine whether the individual to join the group and make a rate or not. The rating of the individual will affect the opinions and the behaviors of subsequent individuals. A simple dynamic model with continuous opinion based on this process has been introduced to predict the number of personnel in online rating. It carries out experiments with the online rating data of films on the Internet website of Douban. The results show that the model can effectively predict the number of online rating; individual final opinion is mainly affected by the opinions of bad-normal-good in the group and almost has nothing to do with its initial opinion; the larger deviation of the Poisson parameter to optimum value(1.25) leads to the lower accuracy of prediction.