资源科学 ›› 2021, Vol. 43 ›› Issue (5): 872-885.doi: 10.18402/resci.2021.05.02

• 资源管理 • 上一篇    下一篇

基于机器学习的雾霾污染精准治理

孙亚男(), 费锦华   

  1. 山东财经大学工商管理学院,济南 250014
  • 收稿日期:2020-03-11 修回日期:2020-08-11 出版日期:2021-05-25 发布日期:2021-07-25
  • 作者简介:孙亚男,男,山东蓬莱人,博士,教授,主要研究方向为经济增长与绿色发展。E-mail: sun_ynan@163.com
  • 基金资助:
    山东省自然科学基金项目(ZR2018MG010);教育部人文社会科学基金项目(19YJCZH107);山东省高等学校人文社会科学计划项目(J17RA080)

Precise governance of haze pollution based on machine learning

SUN Yanan(), FEI Jinhua   

  1. School of Business Administration, Shandong University of Finance and Economics, Jinan 250014, China
  • Received:2020-03-11 Revised:2020-08-11 Online:2021-05-25 Published:2021-07-25

摘要:

实施雾霾污染精准治理是应对局部地区雾霾污染的重要举措。本文利用2001—2016年263个中国城市PM2.5数据,基于机器学习构建了决策树递归分析方法,量化了雾霾污染分区因素和治理因素的交互影响,进而识别出雾霾污染精准治理区域及其治理因素。结果表明:①地理区位、行政级别、产业结构、区域规划、经济地带是识别中国城市雾霾污染精准治理区域的分区因素,据此本文识别出4类城市雾霾污染精准治理区域。②对于沿海城市,提高经济发展水平、降低第二产业占比有利于提高雾霾污染治理水平;对于非沿海城市,降低人口密度有助于缓解雾霾污染,其中非沿海、非省会城市提高科技进步水平也可以有效控制雾霾污染。③中国城市雾霾污染治理分区因素的长期演变趋势与中国“五年规划”期间的国家政策具有高度契合性。因此,科学施策、精准调控有利于雾霾污染精准治理。

关键词: 机器学习, 城市雾霾污染, 精准治理, 影响因素, 递归分析

Abstract:

The implementation of precise governance of haze pollution is an important measure to deal with haze pollution. Using the PM2.5 of 263 Chinese cities from 2001 to 2016, this study constructed a recursive partitioning analysis method of decision tree based on machine learning, quantified the interactions between haze pollution zoning factors and governance factors, and then identified the haze pollution governance regions and governance factors. The results show that: (1) Geographic location, administrative level, industrial structure, regional planning, and economic zone are the main factors to identify the precise governance regions of urban haze pollution in China. Based on these factors, four types of urban haze pollution governance regions were identified. (2) In coastal cities, increasing the level of economic development and reducing the proportion of the secondary industry are conducive to improving the level of haze pollution governance. In non-coastal cities, reducing population density is helpful for alleviating haze pollution. Furthermore, in non-coastal and non-provincial capital cities, improving the level of science and technology progress is beneficial for controlling haze pollution. (3) The long-term trend of urban haze pollution governance zoning factors in China is highly consistent with the national policies during the “Five-Year Plan” periods. Therefore, scientific policies and precise regulation are conducive to the Precise governance of haze pollution.

Key words: machine learning, urban haze pollution, precise governance, influencing factors, recursive partitioning