论文部分内容阅读
目前雾霾污染日益严重,威胁到了环境保护和人类健康,需要对雾霾天气进行预测.通过对多个支持向量机(SVM)进行选择性集成,克服单个SVM不稳定的缺点,提出了基于相异度的SVM选择性集成雾霾天气预测方法(DSE-SVM).首先采用高斯核SVM独立训练出多个个体SVM;其次计算出个体SVM的相异度,剔除相异度最大的个体SVM;最后运用多数投票算法对剩余的SVM进行集成,并进行了理论分析.通过对北京、上海和广州三地区近两年的雾霾数据进行实验分析,实验结果表明DSE-SVM方法预测性能更优,具有较高的稳定性和可信性.
At present, the smog pollution is becoming more and more serious, threatening the environmental protection and human health, and the haze weather needs to be predicted.Through the selective integration of multiple support vector machines (SVMs) and the shortcoming of instability of single SVM, The heterogeneous SVM is selectively integrated with the haze weather forecasting method (DSE-SVM). Firstly, Gaussian kernel SVM is used to train multiple SVMs independently. Second, the dissimilarity of individual SVMs is calculated, Finally, the majority of voting algorithms are used to integrate the remaining SVMs and the theoretical analysis is carried out.The experimental results show that the DSE-SVM method has better performance in forecasting haze data in the past two years in Beijing, Shanghai and Guangzhou, Has a high stability and credibility.