针对目前较严重的雾霾污染, 雾霾天气预报显得十分重要, 通过将改进人工鱼群算法和分形学习相结合, 提出了基于人工鱼群和分形学习的雾霾天气预报方法. 首先对人工鱼群算法离散化改进, 结合分形学习理论降维雾霾数据; 其次运用支持向量机和5-折交叉验证技术分类分布可能不平坦的数据集; 最后建立雾霾天气预报模型. 实验结果表明, 数据降维后更有利于提高分类器性能, 与传统预报方法相比, 预报性能更优, 具有较高的稳定性和可信性.

Haze prediction method based on artificial fish swarm and fractal learning was proposed, by improving artificial fish swarm algorithm combined with fractal learning, because haze forecast is very important for serious haze pollution at present. Firstly, reducing dimension of data through making artificial fish swarm algorithm discrete combined with fractal learning; Secondly, classifying disposing non-flat data set of after dimension reduction by 5-fold cross validation and SVM; Finally, building a haze forecast model. Experimental results show that the data of after dimension reduction is conducive to classify, the prediction method is superior to the traditional methods, and has relatively high stability and credibility.