Stochastic Programming Based Risk Management Of Virtual Enterprise

With the development of global economy, the scheme of virtual enterprise (VE) has been the best choice for modern enterprises to adapt to the competition, due to VE’s "win-win" goal. There are many advantages of VE, such as optimize resource utilization, increase scale of the business and complementary capabilities of the business partners. However, in this paradigm, there are many risks that threaten the security of VE, some of them are unprecedented. This paper analysises the stochastic characteristic of the risks in VE, and then considers the risk management problems of Centralized Decision Making (CDM) and Distributed Decision Making (DDM) based on the stochastic programming theory respectively. The works of this paper will be shown by four parts below.(1) Centralized decision making based multi-action and single-choice risk management.In centralized decision making, members of VE (owner and partners) form a team, the only one decision maker is owner. Many risk factors of VE rise from character of stochastic, which is expressed by random variables, and a stochastic programming model is proposed to describe this. In the model, each risk factor can select only one action or do nothing with it. In order to deal with the stochastic variables in the model, Monet Carlo Simulation (MCS) is combined with genetic algorithm (GA) and particle swarm optimization (PSO) respectively, and then the results from two algortims are compared.(2) Centralized decision making based multi-action and multi-choice risk management.By considering a more complex situation, the multi-action and single-choice risk management model is extended to a multi-action and multi-choice one, in which each risk factor can select multiple actions or do nothing with it. The solving of the model is more difficult with the multiple choices of actions. The Monet Carlo simulation is combined with GA and PSO to solve the model respectively.(3) Constructional distributed decision making (CDDM) based risk management.In distributed decision making, there are different decision makers in different decision levels, such as owner and partners. The main feature of CDDM is the information symmetry between owner and partners, the owner can anticipate the risk status of partners completely. From different view, we propose two CDDM models of risk management.CDDM model of considering partners’ risk level. It is a two-level model, top-level and base-level are used to describe the decision process of owner and partners respectively. The owner allocates bugets amonge memebers of VE, and the partners are required to be in low risk level. In the base level, the partners minimize its risk level under the allocated budgets.CDDM model of considering returns of risk management. To insure a productive risk management, the return of risk management should be in a required level, namely, the return of risk management is no less than total budgets. A two-level model is used to describe the decision process of owner and partners.To solve the two proposed models, the Basic Particle Swarm Optimization (BPSO) is extended to a Multi-swarm Particle Swarm Optimization (MPSO). The multiple swarms are allocated to owner and partners of VE.(4) Organizational distributed decision making (ODDM) based risk management.For ODDM, the private information is analysised, which results the information asymmetry between owner and partners. The owner can not anticipate the whole risk status of partners, and can only anticipate in a certain probability. The uncertain in top-level is expressed by random variables and a stochastic programming model is built. The two CDDM models are extended to two ODDM models of risk management respectively.ODDM model of considering partners’ risk level. It is a two-level model. The top-level describes the decision process of owner, who minimizes the risk level of VE while keeping each partner in low risk level. In the base-level, the decision makers are partners. They have to make decision according to its relatively more information and the owner’s instruction.ODDM model of considering returns of risk management. It is a two-level model, top-level describes the decision process of owner. The owner allocates budgets among members of VE to minimize VE’s risk loss and keep the returns of risk management. In the base-level, partners make decision according to its relatively more information and the owner’s instructionTo solve the two proposed ODDM models, Monet Carlo Simulation is combined with MPSO (MCS-MPSO).