主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2025, Vol. 33 ›› Issue (5): 124-137.doi: 10.16381/j.cnki.issn1003-207x.2024.1770cstr: 32146.14/j.cnki.issn1003-207x.2024.1770

• • 上一篇    下一篇

数据驱动的贫困识别、分类与预测方法研究

谭索怡1, 王梦宁1, 田野2, 刘建国3, 吕欣1()   

  1. 1.国防科技大学系统工程学院,湖南 长沙 410073
    2.清华大学工业工程系,北京 100084
    3.上海财经大学数字经济系,上海 200433
  • 收稿日期:2024-09-30 修回日期:2024-12-03 出版日期:2025-05-25 发布日期:2025-06-04
  • 通讯作者: 吕欣 E-mail:xin.lu.lab@outlook.com
  • 基金资助:
    国家自然科学基金杰出青年科学基金项目(72025405);国家自然科学基金创新研究群体项目(72421002);国家自然科学基金面上项目(72474223);湖南省科技创新计划项目(2024RC3133);国防科技大学基石基金计划(JS24-04)

Data-Driven Models and Applications on Poverty Identification, Classification, and Prediction

Suoyi Tan1, Mengning Wang1, Ye Tian2, Jianguo Liu3, Xin Lu1()   

  1. 1.College of Systems Engineering,National University of Defense Technology,Changsha 410073,China
    2.Department of Industrial Engineering,Tsinghua University,Beijing 100084,China
    3.School of Digital Economy,Shanghai University of Finance and Economics,Shanghai 200433,China
  • Received:2024-09-30 Revised:2024-12-03 Online:2025-05-25 Published:2025-06-04
  • Contact: Xin Lu E-mail:xin.lu.lab@outlook.com

摘要:

贫困是指个体或群体无法获得满足基本生活需求的资源和服务,长期以来一直是全球关注的重大社会问题。传统的贫困识别和测量方法主要依赖于统计数据和抽样调查,存在成本高、效率低、时效性差和数据匮乏等局限性,难以及时反映动态的社会经济状况。随着数字化时代的来临,人口、地理和经济等领域的数据资源日益丰富,为运用人工智能与大数据技术深入研究贫困问题提供了新的契机。本文对贫困的关键概念和测量指标进行了系统回顾,重点探讨了数据驱动的模型和算法在绘制贫困地图、预测贫困趋势和评估社会经济发展状况等方面的应用。在此基础上,进一步讨论了大数据和人工智能技术在实现减贫目标上的潜在影响和机遇,如更精细的空间分析、实时监测能力和趋势预测,同时,也指出了数据可代表性、数据质量和模型可解释性等关键挑战,以及未来可能的发展方向。

关键词: 可持续发展目标, 贫困测量, 大数据, 人工智能

Abstract:

Poverty refers to individuals or groups who are unable to obtain the resources and services necessary to meet basic living requirements, and it has long been a major global social issue. Traditional methods of poverty identification and measurement mainly rely on statistical data and sample surveys, which are limited by high costs, low efficiency, poor timeliness, and data scarcity, making it difficult to reflect dynamic socioeconomic conditions in a timely manner. With the advent of the digital era, data resources in fields such as population, geography, and economy are increasingly abundant, providing new opportunities for the use of artificial intelligence (AI) and data-driven models to tackle poverty in more precise and timely ways. It systematically reviews the key poverty concepts and measurements in this paper, focusing on the application of data-driven models and algorithms in poverty mapping, poverty trend prediction, and socioeconomic status assessment. It is organized as follows: Section 2 provides a succinct overview of the diverse definitions of poverty, and summarizes both unidimensional and multidimensional measurements of poverty through the lenses of education, health, and the living environment, among other perspectives. Section 3 to 7 enumerates the data-driven models found in the existing literature, categorizing them systematically based on the various types of methodologies, including regression analysis, machine learning, neural networks, complex network theory, and natural language processing (NLP). To conclude, the potential implications and opportunities for utilizing big data and AI technologies in achieving poverty reduction goals are discussed in section 8, and the forefront is pointed out as well as critical challenges of the field, such as more precise spatial analysis, real-time monitoring capabilities, and trend prediction. At the same time, key challenges are highlighted such as data representativeness, data quality, and model interpretability, while also pointing out possible future directions.

Key words: SDG, poverty measurement, big data, artificial intelligence

中图分类号: