不确定环境下具有稀疏特征的鲁棒投资组合选择问题研究

项目名称： 不确定环境下具有稀疏特征的鲁棒投资组合选择问题研究

项目编号： No.71501155

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 管理科学

项目作者： 王美花

作者单位： 西安电子科技大学

项目金额： 18万元

中文摘要： 由于金融市场的不确定性，传统的均值-方差最优化模型及算法作为量化投资分析的重要研究技术在量化投资实务中表现出模型过度拟合等缺陷，对不确定环境下具有稀疏特征的鲁棒投资组合选择问题展开研究能够有效解决该关键问题，从而提高投资效率、提升获利能力。本项目基于正则化方法和鲁棒优化方法，以降低交易成本、改善模型解的适定性为出发点，根据市场历史数据及投资者的需求确定参数波动的不确定集，构建鲁棒稀疏的单阶段及动态投资组合选择模型。针对含非凸正则项、交易成本等相关模型的非凸性质，设计快速高效的优化算法，克服已有算法不能有效求解大规模问题的缺陷，满足量化投资产品在线管理的需求。最后利用标准测试数据集及我国量化投资产品的相关数据进行实证研究，检验模型的合理性及算法的有效性。本项目研究成果有助于提高我国在量化投资领域以及大规模基金管理方面的研究水平，提升投资者的获利能力。

中文关键词： 鲁棒优化；正则化方法；稀疏特征；动态规划；量化投资

英文摘要： Due to the uncertainty of the financial market, the traditional mean-variance models and relative algorithms as fundamental technique in quantitative investment analysis have the fault of overfitting Researches on robust portfolio selection problem with sparse characteristics under uncertain environment can be able to solve this crucial issue, and also improve investment efficiency and profitability. This project aims to reduce transaction costs , improve the well-posedness of model’s solution, model the uncertain sets according to historical data sets and requirement of the investment, and then establish sparse robust single-stage and dynamic portfolio selection models based on regularization and robust optimization methods. For the non-convex nature of the models, we aims to design fast and efficient optimization algorithms to overcome the disadvantage of the existing algorithms which can not effectively solve the large-scale problem so as to satisfy the needs of portfolio management. Finally, the empirical research is conducted with standard test data and historical data of the quantitative investment products in China, to examine the model’s reasonableness and algorithm’s validity. The research results help improve our research level in the field of quantitative investment and large-scale fund management, and also enhance the investor’s profitability.

英文关键词： robust optimization;regularization method;sparse characteristic;dynamic progarmming;quantitative investment