大型多期投资组合优化问题研究——用对角二次近似方法-金融学专业论文.docx

PAGE PAGE VII VI VI 摘要 多期随机规划（MSP）是一种非常重要的决策支持工具，它能够帮助我们有 效地解决现代投资组合优化问题。在这篇文章中，我们采用二次对角近似法来解 决大规模多期随机规划问题。首先，我们建立一个常规的采用大规模多期随机规 划方法的优化问题，并给定相应的限定条件。然后，我们采用分解法——即将增 广拉格朗日式分解为 n 个凸子问题。 为了解决最优化问题，我们首先要采用分解近似法来分解增广拉格朗日式， 来将原始问题分解开来；接着，我们重新定义子问题，并通过采用障碍函数法来 解决每个凸子问题。最后，通过实际的数值检验来证明二次对角近似方法（The Diagonal Quadratic Approximation or DQA）能够成功的解决多期随机规划问题。 关键词：多期随机规划、二次对角近似法、投资组合优化 Abstract Multistage stochastic programming (MSP) is a very important decision-support tool which helps professionals to model modern portfolio optimization problems. In this thesis, we develop and test the diagonal quadratic approximation method, in order to seek for solutions of MSP problems in large scale. To start with, we set up an optimization problem in the general form of large scale MSP, which is constrained by corresponding conditions. Then, a decomposition method which split the augmented Lagrangian to n convex subproblems is applied to solve the optimization problem. To achieve this, we first employ a separable approximation for the augmented Lagrangian so that the original problem is separated. Furthermore, we can reformulate the subproblem so that a barrier method may be implemented to solve each convex subproblem. Finally, the real-world numerical experiments confirm that the diagonal quadratic approximation method can successfully achieve solutions of MSP problems. Keywords: MSP, DQA, portfolio optimization 目录 HYPERLINK \l _bookmark0 第一章 引言 1 HYPERLINK \l _bookmark1 1.1 文献综述1 HYPERLINK \l _bookmark2 1.2 论文结构5 HYPERLINK \l _bookmark3 第二章 随机规划 6 HYPERLINK \l _bookmark4 2.1 随机规划简介6 HYPERLINK \l _bookmark5 2.2 情景树（Scenario Tree） 8 HYPERLINK \l _bookmark6 2.3 非预测性变量10 HYPERLINK \l _bookmark7 2.4 金融资产配置最优化问题11 HYPERLINK \l _bookmark8 第三章 二次对角矩阵近似法 13 HYPERLINK \l _bookmark9 3.1 定义问题13 HYPERLINK \l _bookmark10 3.2 增广拉格朗日式14 HYPERLINK \l _bookmark11 3.3 对角二次近似法（DQA） 16 HYPERLINK \l _bookmark12 3.4 公式变形18 HYPERLINK \l _bookmark13 第四章 数值检验 20 HYPERLINK \l _bookmark14