基于粒子群优化的最小二乘支持向量机雾霾预测模型

A Haze Prediction Model based on Particle Swarm Optimization for Least Squares Support Vector Machine

论文通过主成分分析提取影响空气质量的主要因素,建立粒子群优化的最小二乘支持向量机预测模型的方法对石家庄市空气质量指数进行预测。结果表明:对雾霾天气影响因子降维后,主成分之和能够描述大于99%的信息表明主成分分析适用于雾霾影响因子的降维;基于粒子群算法计算得到的最小二乘支持向量机正则化参数和核函数参数分别为30、1.0386;粒子群收敛性能曲线稳定在0.0238左右,动态性能曲线稳定在0.015左右;未来七天的空气质量指数仿真预测结果与实际数值的相对误差平均值为0.01。可见基于粒子群寻优算法的最小二乘支持向量机预测模型在短期雾霾天气预测中具有很好的应用价值。

In order to explore the causes of haze weather and accurately predict the air quality, the main factors affecting air quality are extracted by principal component analysis, and the LS-SVM model of particle swarm optimization is established. The method predicts the air quality index of Shijiazhuang City. The results show that after the haze weather influence factor is reduced, the sum of the principal components can describe more than99% of the information which shows that it can be applied to the dimensionality reduction. The regularization parameters and kernel function parameters of LS-SVM calculated by PSO algorithm are 30 and 1.0386, respectively;the PSO convergence performance curve is stable at 0.0238 and the dynamic performance curve is stable at 0.015;the average relative error between the simulation results and the actual values for the next seven days is 0.01. It is concluded that the LS-SVM model based on particle swarm optimization algorithm has a good application value in short-term haze weather prediction.