摘要
为解决西安市空气质量优化问题中,传统空气质量预测方法的不足问题,提出了一种以筛选后决策树为基学习器的随机森林算法来分析研究空气质量影响因子。方法通过将决策树基分类器进行筛选得到更优秀的分类器并进行结合,从而得到影响空气质量污染气体的贡献性,并对污染气体的污染性进行排名,可以更有针对性地进行空气污染治理。通过对西安市空气数据进行实验验证,以及预测误差及决策树规模关系的分析研究,得出PM10是造成空气污染最严重的气体。从实验的结果可以看出,对PM10的治理可以更有效地改善空气质量。
The traditional air quality prediction methods have the shortcomings in solving air quality of Xi’an city. Therefore, this paper proposes a random forest algorithm to analyze and study the air quality impact factors. Using this method, we obtained better classifiers and combined them by selecting decision tree-based classifiers, thus obtaining the contribution of air pollution gases. Ranking the pollution of polluting gases can effectively carry out the air pollution control. Through the verification of air data in Xi’an, and the analysis and research of the relationship between prediction error and decision tree scale, PM10 is the most serious air pollution gas. The experimental results show that the treatment of PM10 can improve the air quality more effectively.
作者
朱飞鸿
陈文胜
侯帅
揣锦华
ZHU Fei-hong;CHEN Wen-sheng;HOU-Shuai;CHUAI Jin-hua(The Second Monitoring and Application Center,CEA,National Earthquake Data Backup Center,Xi'an Shanxi 710054,China;The 39th Research Institute of China Electronics Technology Group Corporation,Xi'an Shanxi 710065,China;School of Information Engineering,Chang'an University,Xi'an Shanxi 710064,China)
出处
《计算机仿真》
北大核心
2020年第10期221-225,共5页
Computer Simulation
关键词
空气质量
集成学习
随机森林
污染气体
Air quality
Ensemble Learning
Random forest
Pollution gas
