项目名称: 模拟仿真的输入不确定性及其在金融风险管理中的应用
项目编号: No.71201117
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 管理科学与工程
项目作者: 胡照林
作者单位: 同济大学
项目金额: 19万元
中文摘要: 本项目研究金融风险管理中相容风险度量(CRM)的设计与估计。我们从金融机构具体实际出发考虑两种情形。在第一种情形下,我们利用样本数据构造分布参数的置信域。然后假设分布参数在该置信域中变动,来考虑最大期望损失,从而设计好一个对应于该置信域的CRM。我们应用似然比方法把CRM估计问题转化为标准的随机优化问题,继而应用近年来发展成熟的样本平均近似方法,设计一序列凸优化近似的算法来解决该优化问题。在第二种情形下,我们假设概率测度集所包含的分布个数为有限个,并将相应的CRM估计问题转化为一个大规模ranking and selection(R&S)问题。我们运用云计算的思想和框架,结合R&S问题的算法方法来处理所得问题。在云计算的构架下,用以估计CRM的算法设计会有很大自由度,我们将集中研究如何设计和优化算法使得估计更有效。本研究将为金融机构更全面地评估其风险提供强有力的工具。
中文关键词: 模拟仿真;输入不确定性;金融风险管理;仿真优化;风险测度
英文摘要: This project studies the design and estimation of coherent risk measure (CRM) in financial risk management. We consider two scenarios that reveal the real situations of the financial institutes. In the first scenario, we use sample data to construct confidence region for the parameters of the distribution. We then assume the parameters vary in the confidence region and consider the largest expected loss, which naturally induces a CRM corresponding to the confidence region. We apply the likelihood ratio method to convert the estimation of CRM to a standard stochastic optimization problem, and then implement the well developed sample average approximation method and design a sequential convex approximations algorithm to solve the newly formulated optimization problem. In the second scenario, we assume the set of probability measures includes a finite number of distributions, and then treat the estimation of CRM as a large scale ranking and selection problem. We deal with the obtained problem by combining the idea and framework of cloud computing with the algorithms and approaches of ranking and selection problems. Under the framework of cloud computing, designing the algorithms for estimating CRM will admit considerable flexibility, and we will focus on studying how to design and optimize the algorithms to make th
英文关键词: Simulation;Input Uncertainty;Financial Risk Management;Simulation Optimization;Risk Measure