The Message Spreading Dynamic And Control Strategy In Complex Network

Complex network is built on the random realization of a certain law,which can better simulate a series of real systems in human society.The research on the spreading dynamic of complex network contributes to the control of disease transmission,virus infection and rumor spreading.This paper makes a deep research on the aspects of multiple messages spreading,information-based disease control strategy and prediction-based disease control strategy.The innovative achievements are as follows.(1)A novel model for multi-messages spreading over complex networks is proposed.The model is a natural evolution of the classical SIS model,considering the relationship between messages.The correlation between messages tends to affect the dissemination of their both.As a receiver,people tend to distinguish the correlations and differences when meet two different messages,and then decide which to believe.The author discovers the relationship between messages and the level of its influence.The research contributes to the finely spreading control of diseases or rumors,and the better guidance to the public sentiment.(2)A practical method of epidemic control based on quarantine and message delivery is proposed.The disease was divided into two latent period and invasion period.Nodes at the invasion period have a higher isolation rate.Let nodes in quarantine send messages to its directly and indirectly neighboring nodes.The immune capacity in these neighbors will increase with a decreasing rate.The message received in a later time acts with a lower effect than which in the former time.A series of simulation experiments are performed on the classical network and the real data set.Results show that the method can achieve a better control of disease with a lower cost.(3)A more advanced disease control strategy based on neural network and reinforcement learning is proposed,which is an improvement of previous strategy.The neural network is adopted for its good fitting and predicting accuracy for nonlinear data.A reinforcement learning model is proposed to simulate the infection process and give a feedback to the neural network.Both the models are integrated in the proposed SIQS disease spreading model.Finally,a minimal propagation cost is harvested by the real-time parameter training and greedy strategy.