Information Cascade Dynamics On Complex Networks

At present,the existing research can not evaluate the network information cascade comprehensively,and the study of network structure evolution and its dynamic characteristics has become an important topic in the field of complex systems.Exploring the underlying mechanism of information cascade is helpful to understand social group events,financial turmoil and global diffusion cascade events of biological communication,and even to control public opinion orientation in social networks.This paper extends the k-q cascade model to explore the cascade process in complex networks.The model is applied to the cascade failure behavior in single-layer networks and double-layer coupled networks respectively,as well as the cascade behavior in the process of information dissemination on networks.The main work of this paper is as follows:1)Starting from the study of cascade failure of single-layer network,which is a special information cascade,the survivability of nodes is innovatively correlated with the degree of nodes,giving the individual differences of nodes.The survivability of nodes is no longer the same,but is specifically set as positive correlation,negative correlation with degree,and complete randomization.A large number of simulation experiments have found that under the personalized settings,the collapse behavior of the system will change greatly.Among them,the setting of positive correlation with nodal initial degree significantly enhances the collapse dynamics of complex networks,and improves the flexibility of networks,especially for scale-free networks.2)Then we explore the general phenomenon of information cascade and study the cascade effect of double-layer networks.The k-q cascade model is innovatively extended by introducing an independent mapping layer to match the mapping relationship between nodes.The nodes and their mapping nodes are regarded as a minimum unit.Two kinds of node separation conditions are proposed,which are strong logic AND and weak logic OR.It is found that the metastable state still exists in the evolution process of bilevel networks,and the type of networks and the mapping mode of nodes will greatly affect the cascade failure process.In the case of strong logic departure rule,the probability of cascade failure of network decreases gradually with the mapping relationship from positive correlation to random correlation and then to negative correlation.However,in the case of weak logic departure rule,the conclusion is just the opposite.3)k-q cascade model is applied to information diffusion.The traditional threshold model is innovatively redefined.If the traditional threshold only focuses on the proportion of neighbors around a node,the threshold condition is not easy to trigger when a node has many neighbors.In the k-q cascade model,the threshold is not only limited to the proportion of neighbors,but also includes the absolute number of neighbors.That is to say,the condition of deciding whether the information is adopted is not only limited to the threshold,but also considers the absolute number of neighbor nodes,which exceeds the given value k_s,and the nodes will also obey the decision of neighbor nodes.The simulation results show that the information triggered by a stimulus point can also be extended to the whole ER stochastic network in cascade model.Considering the propagation speed,it is found that the propagation speed of information on k-q cascade model is faster than that of traditional threshold model.