随着工业化进程的加剧, 雾霾已严重影响 到人类的日常生活, 分析天气因素进而得出影响雾霾天气的关键因子尤为重要.预测雾霾天气形成的关键因子是一个不断剔除冗余因素保留关键要素的过程, 每一个天气因素都有两种状态, 被选中为关键因子与否, 文章根据该特点, 从一维细胞自动机入手, 提出了一种以二元蚁群算法作为搜索策略, 分形理论作为子集评估度量准则的混合方法.因二元蚁群算法前期信息素匮乏需要较长搜索时间, 引入二元粒子群算法对其进行优化, 将粒子经过多次迭代之后得到的最优位置通过模糊函数映射成蚂蚁所需的信息素, 在较短的时间内形成一条信息素落差明显的路径, 缩短算法前期运行时间.最后将所用方法应用于北京, 广州和上海三地雾霾天气关键影响因子的预测中, 并结合10-交叉验证和SVM算法对预测结果分类准确率进行分析, 通过与其它算法进行对比, 结果表明文章算法预测结果具有较高可信度, 为后期的雾霾治理工作提供了重要的参考依据.

With the development of industrialization in China, air pollution has brought great harm to human daily life, so it is very important to analyze the factors which influence the air quality badly. Essentially, the process of finding the key air pollution influence elements are eliminating the redundant and uninformative as well as noisy factors, then obtaining the key ones, so each weather element has two states: Be selected as a key air pollution factor or not. Therefore, selecting main factors for air pollution is a binary optimization problem actually, and binary ant colony optimization (BACO) combined with Fractal Theory is introduced and applied to solve it. At the initial stage of BACO, due to the deficient of pheromone, the whole ant population need a long time to construct a path with distinct discrepancy level of pheromone, then binary particle swarm optimization (BPSO) is involved to improve the BACO. Furthermore, the datasets of Beijing, Guangzhou and Shanghai are used to conduct experiments, also 10-fold and SVM is involved to analyze the classification accuracy, numerical experiments reveal that our method has higher forecasting accuracy and provides a good tool for environmental improvement.