面向管理决策大数据分析的理论与方法

With the rapid development of emerging technologies in the mobile Internet environment, big data from application fields such as public management, business, finance, and health care continue to arise. This has not only had a profound impact on social and economic life, but is also driving a key revolution in the theory and methodology for managerial decision making and promoting the transition of managerial decision-making research to a big data-driven paradigm. Nevertheless, traditional statistical and econometric methods do not adapt well to the features of big data such as substantial missingness, sparse signals, and high computational complexity, while existing AI methods suffer from poor explainability. Thus, there is an urgent need to develop new techniques for big data analytics as well as foundational and innovative theory. This project will focus on the important application scenarios of managerial decision making, and conduct research on the theory and methods of big data analytics for managerial decision making in five directions: (1) statistical theory and algorithms for distributed data and segmented models; (2) algorithms for missing data imputation in big data; (3) variable screening and dimension reduction for big data; (4) analysis of spatio-temporal panel data with network structure; (5) explainability of AI methods. These five directions are complementary and related to each other, focusing on solving the theoretical and methodological bottlenecks in big data value discovery from different angles, and will provide solid theoretical support and effective solutions for the development of big data-driven managerial decision making in China.

随着移动互联环境下新兴技术的快速发展，来自公共管理、商务金融、医疗健康等应用领域的大数据不断涌现，不仅深刻地改变了社会经济生活的面貌，也孕育着管理决策理论与方法的重大变革，推动管理决策研究向大数据驱动范式转变。然而，传统的统计和计量分析方法无法适应大数据包含大量缺失、信号稀疏、计算复杂性高等特征，现有人工智能方法则受到可解释性不足的制约，亟需发展新的大数据分析技术和基础创新理论。本项目将针对管理决策的重要应用场景，从五个方面对管理决策大数据分析的理论与方法开展研究：（1）分布式与分段模型统计推断理论与算法；（2）大数据中缺失数据的填补算法；（3）大数据变量筛选和降维；（4）带有网络结构的时空面板数据分析；（5）人工智能方法的可解释性。这五个方面既相互独立又相互依托，从不同角度着重解决大数据价值发现中的理论和方法瓶颈问题，将为我国大数据驱动的管理决策发展提供坚实的理论支撑和有效的解决方案。

随着移动互联环境下新兴技术的快速发展，来自公共管理、商务金融、医疗健康等应用领域的大数据不断涌现，这不仅深刻地改变了社会经济生活的面貌，也孕育着管理决策理论与方法的重大变革，推动管理决策研究向大数据驱动范式转变。然而，传统的统计和计量分析方法无法适应大数据包含大量缺失、信号稀疏、计算复杂性高等特征，现有人工智能方法则受到可解释性不足的制约，亟需发展新的大数据分析技术和基础创新理论。本项目针对管理决策的重要应用场景，从五个方面对管理决策大数据分析的理论与方法开展研究：（1）分布式与分段模型统计推断理论与算法；（2）大数据中缺失数据的填补算法；（3）大数据变量筛选和降维；（4）带有网络结构的时空面板数据分析；（5）人工智能方法的可解释性。这五个方面既相互独立又相互依托，从不同角度着重解决大数据价值发现中的理论和方法瓶颈问题，将为我国大数据驱动的管理决策发展提供坚实的理论支撑和有效的解决方案。