融合粗糙集和二元萤火虫算法的雾霾关键影响因素预测方法  

Rough set combine with binary glowworm swarm optimization for key haze influence factors

作　　者：程美英[1,2] 倪志伟[1,2] 朱旭辉[1,2] 

机构地区：[1]合肥工业大学管理学院,合肥230009 [2]教育部过程优化与智能决策重点实验室,合肥230009

出　　处：《系统工程理论与实践》2017年第1期241-252,共12页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(71271071;71301041);国家自然科学基金重点项目(71490725);国家"863"云制造主题项目(2015AA042101)~~

摘　　要：雾霾对人类的日常生活带来极大的危害,因而分析产生雾霾的关键影响因素尤为重要.针对目前传统算法预测雾霾关键影响因素存在的缺陷,从一维细胞自动机入手,提出了一种以基于群落弱连接机制的二元萤火虫算法(CWLBGSO)为搜索策略,粗糙集为评价准则的混合方法.CWLBGSO基于自然界中萤火虫间协同进化的"弱连接"机制,划分搜索空间,为每个子空间分配相应的种群,各子种群中的次优个体相互交互产生新个体,从而保持种群的动态多样性,然后将CWLBGSO结合粗糙集,应用于北京,广州和上海三地雾霾关键影响因素的预测中,并结合10交叉验证和SVM算法对预测结果分类准确率和影响因素进行分析,通过与其它算法进行对比,结果表明本文算法能有效剔除冗余因素,预测结果具有较高的稳定性和可行性.Haze has brought great harm to human daily life, so it is very important to analyze the factors which influence the haze badly. Starting from one-dimensional cellular automata （CA） and the drawbacks of the traditional method, a novel BGSO algorithm with weak-link coevolution mechanism （CWLBGSO） combine with rough set is introduced in this paper. In CWLBGSO, the whole search space was divided into several sub-spaces, and each sub-space has a subpopulation, then after several iterations, suboptimum in each subpopulation will perform crossover operation to keep the dynamic diversity. After that CWLBGSO combined with rough set is applied to forecast the key factors which influence haze badly. The datasets of Beijing, Guangzhou and Shanghai are used to conduct experiments, also 10-fold and SVM is involved to analyze the classification accuracy and influence factors, the experimental results show that our method can effectively eliminate redundant factors, also has relatively higher stability and credibility.

关 键 词：雾霾 一维细胞自动机 二元萤火虫算法 弱连接机制 粗糙集 SVM 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]