管理决策中的分布鲁棒优化  

Distributionally Robust Optimization in Management Decision

作　　者：章宇 张静[1] 刘文欣 李宁心 周冰洁 贾晨诗 寇纲[1,2] 唐加福 Yu Zhang;Jing Zhang;Wenxin Liu;Ningxin Li;Bingjie Zhou;Chenshi Jia;Gang Kou;Jiafu Tang(School of Business Administration,Southwestern University of Finance and Economics,Chengdu 611130;Xiangjiang Laboratory,Changsha 410205;School of Management Science and Engineering,Dongbei University of Finance and Economics,Dalian 100190)

机构地区：[1]西南财经大学工商管理学院,成都611130 [2]湘江实验室,长沙410205 [3]东北财经大学管理科学与工程学院,大连100190

出　　处：《中国科学基金》2024年第5期784-793,共10页Bulletin of National Natural Science Foundation of China

基　　金：国家自然科学基金项目(72371204,72293563,71910107002,71901180);四川省自然科学基金项目(24NSFSC6232);西南财经大学“光华英才工程”的资助。

摘　　要：《中华人民共和国国民经济和社会发展第十四个五年规划和2035年远景目标纲要》要求大力发展物流、交通、医疗等行业,其中做好管理决策是关键。作为管理决策的引擎,分布鲁棒优化以其决策稳健性、计算便利性、应用普适性等突出优势,在近20年得到了学界和业界的广泛关注。本文从优化问题环境参数的不确定性建模、分布鲁棒机会约束优化以及分布鲁棒线性优化等方面,梳理了分布鲁棒优化的经典结论与研究现状,并以车辆路径优化、共享出行车辆调配、医疗手术室调度三个典型问题为例,回顾了分布鲁棒优化在管理决策中的应用。立足大数据时代,当前研究趋势包括:特征数据驱动的分布鲁棒优化、结合具体应用场景的数据驱动型分布鲁棒优化以及数据驱动型分布鲁棒离散优化的算法设计等。这些重要问题的解决将推动该领域的数据化发展,并进一步为服务国家战略赋能。Outline of the 14th Five-Year Plan(2021-2025)for National Economic and Social Development and the Long-Range Objectives Through the Year 2035 mandates substantial development in areas such as logistics,transportation,and healthcare,in which management decision plays a key role.As the engine for management decisions,distributionally robust optimization,due to its prominent advantages in decision robustness,computational convenience,and widespread applicability,has garnered significant attention from academia and industry over the past two decades.This paper systematically reviews the classical results and current research status of distributionally robust optimization,focusing on:(1)uncertainty modeling,(2)distributionally robust chance-constrained optimization,and(3)distributionally robust linear optimization.Taking vehicle routing,fleet management in shared mobility,and healthcare surgery scheduling as examples,this paper then reviews the applications of distributionally robust optimization in management decisions.In the big data era,future research directions include:(1)feature data-driven distributionally robust optimization,(2)application-specific data-driven distributionally robust optimization,(3)algorithm design for data-driven distributionally robust discrete optimization.These directions aim to propel its data-driven development and further empower national strategic initiatives.

关 键 词：管理决策 分布鲁棒优化 分布鲁棒机会约束优化 分布鲁棒线性优化 

分 类 号：C934[经济管理—管理学]