基于小波分解和最小二乘支持向量机的大气臭氧含量时间序列预测

基于小波分解(WT)和最小二乘支持向量机(LSSVM)理论,建立了将二者相结合的大气臭氧含量时间序列预测模型。采用香河等4个观测站的月平均臭氧总量观测样本,经小波分解为不同频段的子序列,将这些子序列分别进行LSSVM预测,最后经小波重构得到月平均臭氧总量时间序列预测结果。实验表明该方法能有效预测大气臭氧含量,与支持向量机(SVM)以及人工神经网络(ANN)的预测结果相比,该方法具有较高的预测精度。

Combination of Wavelet Decomposition and Least Square Support Vector Machine to Forcast Atmospheric Ozone Content Time Series

The atmosphere ozone content forecast model was established based on the combination of wavelet decomposition and advanced Least Square Support Vector Machine (LSSVM) regression. This can be approached in three steps: (1)The observations were decomposed into several different frequency signal subsets,(2)the independent prediction models of decomposed signals with Takens delay embedding theorem and Least-Squares Support Vector Machine (LSSVM) were set up, (3)independent predicted results were integrated as the final prediction with wavelet reconstruction. Application experiments with data from Xianghe and the other three observation stations show that the method can make better prediction effectively for the atmospheric ozone content, as compared with conventional Support Vector Machine(SVM) and Artificial Neural Network(ANN).