计算实验金融工程:大数据驱动的金融管理决策工具  

Agent-based Computational Financial Engineering:A Financial Management Decision-Making Tool Driven by Big Data

作　　者：张维[1] 林兟 康俊卿[2] 熊熊[1] 张永杰[1] Zhang Wei;Lin Shen;Kang Junqing;Xiong Xiong;Zhang Yongjie(College of Management and Economics,Tianjin University;Lingnan College,Sun Yat-Sen University)

机构地区：[1]天津大学管理与经济学部 [2]中山大学岭南学院

出　　处：《管理世界》2023年第5期173-187,共15页Journal of Management World

基　　金：国家自然科学基金项目(基金号:91846000、71790594、92146006)的资助。

摘　　要：在大数据时代,金融系统更清晰地呈现出由其微观单元之间的适应性交互形成的复杂系统内在本质。但传统金融理论受限于数学可解性,常利用理性经济人等假设简化金融系统的一些微观因素,致使理论模型难以捕捉诸多真实金融异象和风险事件,制约了理论的发展及其工程化应用。而今,金融系统的微观因素相关研究取得了重大突破,使得从微观主体异质行为及其交互出发的“自底向上”建模、以及相应的新型管理工具成为可能。这些研究推动了由大数据驱动的计算实验金融学、以及基于此的“计算实验金融工程”管理工具的发展。本文通过建立具有包容性的形式化概念模型,对传统金融和计算实验金融的逻辑和方法论进行阐述和比较,尝试在两者之间搭建对话桥梁。同时在上述模型框架下,通过示例展现了大数据驱动的计算实验金融如何拓宽金融科学理论的发展,并为金融实践中的管理决策带来新工具。In the era of big data,the financial system presents the inherent nature of complex systems that consist of adaptive interactions among the microscopic units that make up the system in a more explicit way.However,due to the limitation of mathematical tractability,traditional financial theory always makes assumptions,such as the representative rational agent,to simplify the essential micro-factors of the financial system,which makes theoretical models encounter difficulties in capturing many financial anomalies and risk events,and further,restricts the development of financial theory and its engineering application.Nowadays,new"bottom-up"modeling methods,which based on the heterogeneous behavior of microscopic subjects and their interactions,and management decision-making tools have become possible due to the unprecedented breakthroughs in the research on the micro-factors that constitute financial complexity.These studies have extensively promoted the progress of Agent-based Computational Finance(ACF)as well as the big data-driven Agent-based Computational Financial Engineering(ACFE)based on Agent-based Computational Finance.By establishing an inclusive conceptual model IBM(Information and Behaviors in Market),this paper elaborates on and compares the logic and methodology of traditional finance and Agentbased Computational Finance by establishing an inclusive formal conceptual model and tries to build a dialogue bridge between the two research frameworks.At the same time,through examples,this paper shows how big data-driven ACF broadens the development of financial theory and how ACFE brings new tools for management decision-making in financial practice.

关 键 词：大数据 金融管理决策 计算实验金融工程 微观行为 复杂异质交互 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] F832[自动化与计算机技术—计算机科学与技术]