不同风险偏好下虚拟企业风险规划研究

Risk Programming for Virtual Enterprise Based on Various Risk Preferences

针对虚拟企业风险规划问题,在分析其各种风险具有随机性的特点的基础上,运用随机规划理论,分别建立风险规划的期望值模型和机会约束规划模型来描述决策者在不同风险偏好下的决策行为。针对所建立的模型,分别设计了基于蒙特卡罗模拟的粒子群优化算法、遗传算法和蚁群算法对其进行求解。仿真分析表明期望值模型较好地描述了风险中性决策者的决策行为,机会约束规划模型随着其偏好系数取值的不同描述了不同风险偏好(风险厌恶、风险中性、风险爱好)决策者的决策行为。通过对三种算法仿真结果的比较分析,表明基于蒙特卡罗模拟的粒子群优化算法在寻优能力、稳定性和收敛速度等方面优于其余两种算法,是解决此类风险规划问题的有效手段。

For the stochastic characteristics of each risk for risk programming of virtual enterprise, with the sto- chastic programming theory, an expected value model and a chance constraint programming model are proposed to describe the decision behavior under various risk preferences. A Monte Carlo Simulation based Particle Swarm Optimization （ MCS-PSO）, Genetic Algorithm （ MCS-GA ） and Ant Colony Optimization （ MCS-ACO ） are designed to solve the models respectively. The simulation analysis shows that the expected value model describes the deci- sion behavior of risk-neutral decision maker, and the chance constraint programming model describes the decision behavior of decision makers with various risk preference（risk-neutral, risk-averse, risk-loving）while the prefer- ence coefficient has different values. The comparison of the simulation results from the three proposed algorithm shows that the MCS-PSO performs better than MCS-GA and MCS-ACO on searching ability, reliability and con- vergence speed, and MCS-PSO is an effective way to solve this kind of risk programming problems.