模糊聚类支持向量机的区域空气PM_(2.5)浓度预报

Fuzzy Clustering Support Vector Machine for Predicting Regional PM_(2.5) Concentration

在分析模糊C均值聚类算法与支持向量机回归的特点后,将二者结合,提出了模糊聚类支持向量机回归(FCM-SVR)算法,对空气中颗粒物浓度PM2.5进行预测.该方法首先利用模糊C均值聚类算法把一个复杂的数据集分成多个群体,再在每个群体上建立支持向量机回归(SVR)模型,然后进行集成,对区域空气的PM2.5浓度进行预测.预测结果分别与自组织竞争神经网络支持向量机回归(SOM-SVR)模型和单一的支持向量机回归(SVR)的结果进行比较.结果表明,FCM-SVR模型的预报准确率高于SOM-SVR模型和SVR模型.

A fuzzy clustering support vector machine regression algorithm is proposed by analyzing and combining the characteristics of the fuzzy C-mean clustering algorithm and the support vector machine regression. The SVM is designed to forecast the particles density PM2.5 in the air. Firstly, a complex data set is separated and inserted into multiple groups using fuzzy C-mean clustering algorithm. Then the SVM regression model in each group is established. The integrated fuzzy clustering SVM regression is applied to forecast the PM2.5 in the local air. By comparing the predicted result with that of the self-organizing competitive neural network SVM regression model, as well as that of the single SVM regression model respectively, it is found that the predicted accuracy rate of the FCM-SVR is higher than that of the SOM-SVR model and SVR model.