大气环境分类的SVM联合决策方法  

SVM Joint-decision Method for Atmospheric Environment Classification

作　　者：王竟成 张伦武[1] 杨小奎[1] 胡学步[2] 周俊炎 李泽华[1] 吴帅[1] WANG Jing-cheng;ZHANG Lun-wu;YANG Xiao-kui;HU Xue-bu;ZHOU Jun-yan;LI Ze-hua;WU Shuai(CSGC Key Laboratory of Ammunition Storage Environmental Effects,Southwest Technology and Engineering Research Institute,Chongqing 400039,China;College of Chemistry and Chemical Engineering,Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]西南技术工程研究所弹药贮存环境效应重点实验室,重庆400039 [2]重庆理工大学化学化工学院,重庆400054

出　　处：《装备环境工程》2022年第8期148-154,共7页Equipment Environmental Engineering

基　　金：中国博士后基金项目(231676)。

摘　　要：目的实现城市大气环境的精准快速归类预测。方法基于支持向量机(SVM)构建多分类问题的联合决策算法,将大量城市环境因素数据的主成分聚类结果作为输入,通过机器学习训练,组建大气环境的SVM联合决策模型。结果该模型根据大气环境因素将数据集91个城市划分为9类,其中河内与海防环境相似度最高,巴东与格尔木差异最大。9个SVM二分类器组建的联合决策模型通过逐点预测在主成分数据空间形成了分区预测云图。结论SVM联合决策模型可实现城市环境的快速分类辨识,分类预测结果精度高于95%。The paper aims to conduct a quick accurate prediction for atmospheric environment classification of different cit-ies.SVM is used to construct a joint-decision algorithm for multi-classification problems,and the principal component cluster-ing results of a large number of urban environmental factor data are input.Through machine learning training,the SVM joint-decision model of atmospheric environment is constructed.In the 91 cities,Hanoi and Haiphong are the most similar cou-ple,while Padang and Golmud turn out to be the most different cities.The joint-decision model formed by 9 SVM binary classi-fiers forms a partitioned prediction cloud image in the principal component data space by point prediction.Results show that the prediction accuracy is higher than 95%,therefore atmospheric environment of different types can be recognized swiftly by the established model.

关 键 词：环境分类 支持向量机 层次聚类 主成分分析 联合决策 机器学习 

分 类 号：P412[天文地球—大气科学及气象学]