基于高性能计算的养老目标基金投资策略研究

In a highly uncertain financial market environment, scientific asset allocation and fund investment strategies are particularly critical for pension funds to invest, so designing dynamic asset allocation and simulating large-scale complex scenario investment strategies scientifically based on high-performance computing technology will notably improve the investment management level and decision-making ability of Target Date Funds, which will be of great significance to the institutional advantage of the accumulated pension mode and to the improvement of the Chinese pension system. In this project, we shall firstly model the selection problems of investment strategies for China's Target Date Funds reasonably, then study the effective algorithms to solve the models, and finally explore and implement the parallel algorithms. Our research will focus on the parallel algorithms for large-scale complex scenarios simulation of dynamic asset allocation in the face of TDFs investment strategies. We shall optimize the computational performance in multi levels and achieve the goals of almost real-time response to the large-scale complex scene simulation of the dynamic asset allocation of Target Date Funds investment strategy, output result in the model toolbox, high-performance algorithmic library and solvers with independent property rights. This project is motivated by the specific financial applications and combined with multiple disciplines. It captures the practical computational bottlenecks of China's Target Date Funds. The innovations of our research lie in the following aspects: Target Date Funds theoretical models, the simulation of large-scale complex scenarios of dynamic asset allocation, and the design and implementation of high-performance global parallel numerical algorithms. Our research will promote the development of related disciplines, the innovation of methods for simulation of financial large-scale complex scenarios and the amalgamation of computational technology & the investment management practice. And this research also owns great practical significance and long-term application value for expanding the industrial applications of high-performance computing and improving the investment management ability of financial investment institutions.

在高度不确定金融市场环境下，科学的资产配置和投资策略对养老基金参与市场投资尤为关键，基于高性能计算技术设计动态资产配置和大规模复杂场景模拟,将极大提升养老目标基金(TDFs)投资管理水平和决策能力,对发挥积累性养老模式的制度优势、完善中国养老体系意义重大。本项目针对TDFs投资策略选择问题，构建中国TDFs理论模型，设计相应的算法并研究其并行化策略；着重研究TDFs投资策略动态资产配置的大规模复杂场景模拟并行算法；利用并行计算技术多层级优化算法性能，实现模拟快速响应，产出具有自主知识产权的TDFs投资策略选择模型工具集、算法库和求解器。该研究由具体应用驱动，多学科交叉，瞄准应用瓶颈，在TDFs理论模型、动态资产配置大规模复杂场景模拟及并行算法设计与实现三方面将有创新。预期成果将促进TDFs理论发展、金融大规模复杂场景模拟方法创新、高性能计算应用与投资管理实践的融合，具有现实意义和应用价值。

在高度不确定金融市场环境下，科学的资产配置和投资策略对养老基金参与市场投资尤为关键，基于高性能计算技术设计动态资产配置和大规模复杂场景模拟，将极大提升养老目标基金（TDFs）投资管理水平和决策能力，对发挥积累性养老模式的制度优势、完善中国养老体系意义重大。本项目针对中国养老基金资产投资管理所面临的资产最优配置以及计算瓶颈问题，研究具有中国特色的养老目标基金理论模型、资产配置路径设计与算法、大规模复杂场景高性能模拟与实现技术。通过本项研究，主要在以下三个方面取得重要进展：（1）养老金投资组合的基础研究。开展了养老金投资组合基础理论研究，设计业绩比较基准方法以及 TDFs 风险与绩效分析方法；（2）养老金投资组合的模型研究。充分考虑养老金投资管理的资产类别等现实约束，合理构建了符合中国实际国情、具有中国特色的TDFs理论模型；（3）养老金投资组合模型的高性能计算研究。设计了中国TDFs理论模型、动态资产配置场景模拟问题的串行求解算法与并行策略，并基于高性能计算环境实现了问题的并行优化。依托本项研究在国内外有影响的学术期刊或会议上共计发表论文18篇，申请专利3项、软件著作权6项，完成会议报告19篇，获得1项科研奖励，举办、参加学术交流共7场。产出多项具有自主知识产权的TDFs投资策略选择模型工具集、算法库和求解器，并将相关成果部署在自有高性能集群可供使用。项目主要参与者韩子栋与陈逸东参加2022第三届中国国际并行应用挑战赛（IPCC），分别获得全国总决赛一等奖与三等奖，组队参加2022全国并行应用挑战赛（PAC）获得全国二等奖，在并行应用方面实现探索与突破。