项目名称: 两类投资组合优化问题的模型与算法研究
项目编号: No.11301041
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 戴志锋
作者单位: 长沙理工大学
项目金额: 22万元
中文摘要: 如何减轻参数扰动对投资组合问题最优权重的影响,是近年来金融和优化领域共同关心的问题。鲁棒优化与正则化是两种解决该问题的主要方法。但鲁棒投资组合优化模型因过于忽略历史信息的部分可知性,存在过于保守的问题。正则化投资组合问题优化尚处于起步阶段,模型还需要完善,且该类问题的求解尚缺少有效的算法。 本项目旨在研究:1)考虑抽样样本的误差,根据资产分布和收益的特征, 构建非对称不确定集,建立新型鲁棒风险价值(VaR)及条件风险价值(CVaR)投资组合优化模型,来解决以往鲁棒模型过于保守的问题。2) 根据正则化投资组合优化问题的特殊结构,设计基于光滑函数的增广拉格朗日乘子法,以及近似梯度下降算法。通过数值模拟和实证分析,检验模型和算法的有效性。 本项目的研究在理论上将进一步丰富现代投资组合理论,在实践中将为投资者提供决策技术参考,也可为最优化理论及算法的应用探索新的方向。
中文关键词: 投资组合;鲁棒优化;正则化;光滑化函数;近似梯度下降法
英文摘要: How to alleviate the parameter perturbation on the optimal weights of portfolio problem, in recent years is an active issue in the field of financial and optimization. Robust optimization and regularization are two main ways to solve this problem. But robust optimization for ignoring partially known historical information is too conservative. Regularization portfolio optimization problem is still in the initial stage. The model also needs to improve. And there is a lack of effective algorithm to solve this kind of problem. This project aims to study: 1) Consider the sampling error of the sample, according to the asset allocation and income characteristics, we build asymmetric uncertainties set and establish a new robust value-at-risk (VaR) and conditional value at risk (CVaR) investment portfolio optimization model in order to solve the problem that the robust model in the past is too conservative. 2)Through studying the special structure of the regularization of portfolio selection, we design smooth function augmented Lagrange multiplier methods and the approximate gradient descent algorithm for it. Through the numerical simulation and empirical, test the efficiency of the models and algorithms. The study of this project, in theory, will further enrich the modern portfolio theory, in practice, will pr
英文关键词: Portfolio;Robust optimization;Regularization;Smoothing function;Proximal-gradient descent method