针对区域PM2.5浓度预报这一问题进行了研究,通过结合支持向量回归机(SVR)和粒子群优化算法(PSO),提出了一种加权因子的预报方法(W-PSO-SVR).该方法采用了对预报模型的输入变量进行[0,1]间的不均等加权赋值,权重值由PSO搜索求得,通过不断寻优迭代,赋予输入变量的不均等权重,从而建立预报模型.采用该方法的区域空气的PM2.5浓度预报结果表明,与单独的支持向量回归机模型和0或1的加权因子的支持向量回归模型相比,W-PSO-SVR预报精度提高明显,能较好地实现对模型输入参数的有效选择.%This paper developed a regional air PM2.5 concentration predicting model with weighting factors (W-PSO-SVR),which combined support vector regression(SVR) and particle swarm optimization (PSO).The [0,1] unequal weighting factors which were achieved by the PSO search were assigned to the input variables of the model.When the unequal weighting factors were confirmed,then it established the PM2.5 predicting model.Compared with the pure SVR model and 0 or I weighting factors' SVR model,predicting results indicate that W-PSO-SVR model performs better and the predicting accuracy is higher.Besides,the W-PSO-SVR model can achieve the better effective selection of input parameters.
展开▼
机译:Assessing Correlation between PM2.5 and Meteorological Variables and Projecting the Impact of Climate Change on PM2.5 =PM2.5与气象因子的相关性分析及气候变化对PM2.5的影响