不确定供应链管理网络的概率图模型仿真研究  

Research on Simulation of Probabilistic Graphical Models in Uncertain Supply Chain Management Network

作　　者：彭青松[1] 张明[1] 叶爱兵[1,2] 

机构地区：[1]上海海事大学信息工程学院,上海200135 [2]上海海事大学物流工程学院,上海200135

出　　处：《系统仿真学报》2008年第S1期307-309,共3页Journal of System Simulation

基　　金：上海市教委支出预算研究项目(2008102)

摘　　要：供应链管理网络是种有向无环图,在实际应用中不可避免的具有多种不确定性,这些不确定性可用概率进行表示。贝叶斯网是种应用较广的概率图模型,它也是有向无环图。对供应链管理网络中的不确定信息与贝叶斯网进行研究,对使用贝叶斯网络研究供应链管理的可行性进行了论证。由于贝叶斯网是概率分布与图表示的完美结合,根据供应链管理中的不确定信息,可建立合适的贝叶斯网,通过联合概率分布进行决策,使得供应链中的各方达到共同获利的目的。由于贝叶斯网在一般情况下的学习与推理问题都是NP难问题,对用于供应链管理贝叶斯网在仿真中可能遇到的情况进行了分析。研究表明,将贝叶斯网应用到供应链管理中的不确定信息仿真是切实可行的。Supply chain management network is a kind of undirected acyclic graph,and has unavoidable uncertainties in real applications,which can be expressed by probability expressively.While Bayesian Network is a typical kind of probabilistic graphical model,which is also a kind of undirected acyclic graph.By using joint probability distribution and graphical model, Bayesian networks can be constructed according to the uncertain information in supply chain management network,and the decision can be made by computing the joint probability distribution to ensure all entities in the supply chain can gain profits. The uncertain information of the supply chain management network and Bayesian network are investigated as well as the conclusion is drawn that applying the Bayesian network to supply chain management network is feasible.The complexity of learning and inference of Bayesian networks being NP-hard problem,some practical issues shall be encountered during the simulation process,which is analyzed in detail.Research result shows that applying the Bayesian network to the supply chain management network is feasible to handle the uncertain information.

关 键 词：贝叶斯网 供应链管理网络 不确定信息 有向无环图 概率图模型 计算机仿真 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]