The Research Of Logistics Network And Node Importance Based On Exponential Random Graph Models

With the rapid development of science and technology,enterprise competition is increasingly fierce.Coupled with the speedy development of Internet E-business,logistics has gradually become the focus of all walks of life.At the same time,traditional logistics activities are costly,time-consuming and complex,which have been far from the requirement of modern enterprise development.Here comes the concept of logistics network--the functions of the logistics activity,integration of resources and information.In the complicated and changing situation,the complexity,nonlinear and multi-modular characteristics of logistics network have become increasingly apparent.The logistics network and the complex network have common behavior patterns,so that applying the theories and ideas of complex network to logistics network researches has become a new hotspot in logistics management.However,using the statistical characteristics and evolution mechanism of the network to study the structure of logistics network seems to have stalled,therefore we need a new perspective to explore the inner mechanism of the logistics network.In this paper,we combine the complex network theory with statistics model to study the weight,structure and node importance of logistics network.The main research work and innovations are as follows:Firstly,a simulation method for the weight of logistics network was proposed,which was named node flow model.Based on the thought of considering the logistics quantity as logistics network edge weight,our passage applies gravity model to simulate the logistics quantity between cities.Through the hypothesis of the relationship between traffic and node degree,we verify the feasibility of the model in theory and practice.Secondly,ERGMs was improved according to the characteristics of logistics network.The improved model combines ERGMs and p₂ model,taking into account the local structure variables of the network and nodal random effects synchronously.In addition,this passage propose the mixed valued ERGMs,where we integrate the weight and the binary network by finding the probability of edge weight.Plus,we estimate the parameters by exerting Bayesian estimation accompany with MCMC algorithm and compare the model with the traditional ERGMs to prove the superiority of the mixed valued ERGMs in parameter acceptance,goodness of fit and Bayesian factor.Thirdly,a ranking algorithm of node importance was come up with in weighted network.Inspired by the nonadjacent nodes and flows based on the propagation path,with the node flow model of the second chapter and the random effect of the third chapter,we extend the existing node importance sorting algorithm.According to the comparison between this algorithm and centrality node importance sorting method,we could draw a conclusion that our method is more consistent with practical situation and more suitable for logistics network.