Agri-food Supply Chain Modeling And Risk Propagation Based On Complex Network Theory

As a necessity of life,agri-food plays an important role in everyone's life.However,the emergence of oversupply,price fluctuation,quality and safety issues cause broad attention of the price stability and safety of agri-food.Supply chain network has been proved to be a typical complex network.This thesis aims to reveal evolution mechanism of agri-food supply chain,and learn risk propagation mechanism to find effective control strategy from the perspective of complex network.Key problems in this topic are:(1)How to develop an evolving complex model that reflects the characteristics of agri-food supply chain?(2)How to describe the topological characteristics of the network?(3)How to characterize the risk propagation mechanism on agri-food supply chain network?(4)How to mitigate and control the risk propagation on agri-food supply chain network?To address these problems,the main content of this thesis are as follow:1.Analyzing the structure and characteristics of agri-food supply chain in China.At present,wholesalersare core members in most agri-food supply chains.The household-based peasant economy leads to prominent contradiction between “small farmers” and the “big market”.The most typical agri-food supply chain consists of farmers,dealers,wholesalers,retailers and consumers in China.All these factors lead to a multi-level,complete competition,information asymmetry and unstable supply chain network.2.Developing the evolving models of agri-food supply chain network.Considering a four-layer agri-food supply chain network that consist of farmers,dealers,wholesalers and retailers,a dual-local world fitness model is built to generate an undirected and unweighted network.Furthermore,a directed and weighted network is modeled based on the balance of volume.The dual-local world based fitness model(DLWF model)considers two characteristics in supply chain:(1)the local world is different for different kinds of node rather than randomly selected;(2)the distribution of fitness is determined by its physical meaning,which is the comprehensive attraction for edgeing other nodes.Both normal distribution and power law distribution are considered to describe the different situations of nodes.When fitness is normally distributed,edges can only be connected between nodes of adjacent layers;while fitness is power law distributed,edges can be connected among nodes of different layers.The distribution of network degree shows a generalized free-scale network with variable power exponent,which determined by the distribution of fitness,proportion of different kinds of nodes and size of local world.On the basis of previous DLWF model,a directed and weighted model is built considering the direction and balance of transaction between nodes,which called DLWF Model for weighted networks(DLWF-W model)in this thesis.The distribution of edge strength,vertex strength and degree all show generalized free-scale network with variable power exponent.For each vertex,it has a different power exponent,which is determined by the category of node,distribution of fitness,and proportion of different kinds of nodes.3.Analyzing the topological characteristics of agri-food supply chain network by simulation.Simulation results of degree distribution,edge strength and vertex strength show the same rule as theoretic analysis.They demonstrate a power law tail with different power exponents in a log-log coordinate.The network has a long average supply chain path length and small clustering coefficient.The results also show that when the fitness follows a power law distribution,the average supply chain path length decrease and the clustering coefficient increase because edges can be generated between vertexes from nonadjacent layers.4.Studying the risk propagation dynamics on agri-food supply chain complex network.According to their propagation mechanism and consequences,supply chain risks can be classified into two categories: quality risks and supply-demand risks.For quality risks,its propagation dynamics is studied by using SIR(susceptible,infected,removed or recovered)epidemic model.Considering the decrease of the risk spreading rate with risk exposure and countermeasures,this paper introduces an inhibiting factor to the spreading rate to classical SIR model.Thus,a SIR model with non-linear spreading rate is developed to analyze the propagation mechanism.The threshold of risk spreading rate and node influence are worked out by differential equation and then applied to the agri-food network generated by DLWF model.The result shows that although the inhibiting factor affects the speed of risk spreading,it has no effect on the propagation threshold rate and range of risk spreading.The risk propagation threshold is 0 when the network degree distribution power exponent is between 2 and 3;while the power exponent is greater than 3,the network has a positive threshold,and the propagation threshold is close to that of random network as the exponential increases.For supply and demand risks,they are related to node capacity and can be described by the load-capacity model.The author employs a load-capacity on node dynamics with considering the physical meaning of load in agri-food supply chain network:(1)initial load represents the transaction volume for a node;(2)a node will refuse those volume when the supply exceed its capacity.The threshold of cascading failure for DLWF-W model is calculated as a piecewise function,and determined by network structure and load redistribution principle.5.Strategies of risk propagation on agri-food supply chain network.Immunization strategy and structure optimizing are suggested to mitigate and control the risk propagation and cascading failure.By comparing the effects of random immunization,target immunization,acquaintance immunization,and category immunization,we fond that target immunization is the most effective strategy,while acquaintance immunization is a cost-effective strategy and random immunization is ineffective on agri-food supply chain network.For structure optimization of the network,it is helpful to improve the fitness of farmers and retailers.From network perspective,larger fitness of farmers also improves the robustness of the network by changing the proportion of different nodes,reducing average supply chain length,and increasing clustering coefficient.In practice,developing cooperatives,farmer market,chain stores,direct sale from cooperatives to supermarkets,these are all helpful to increase efficiency and improve the capability of agri-food chain against risks.This study has both theoretical and practical value.This thesis enriches the research of complex network in supply chain management.Agri-food supply chain network modeling is crucial to reveal the formation and evolution mechanism of the network.The analysis of risk dynamics provides practical strategies for risk control on the network.