基于小波分解和最小二乘支持向量机的西太平洋副高预测模型

AREA EXPONENT OF WESTERN PACIFIC SUBTROPICAL HIGH FORECAST MODEL BASED ON WAVELET DECOMPOSITION SUPPORT VECTOR MACHINE

用小波分解(WT)和最小二乘支持向量机(LS-SVM)相结合的方法,建立西太平洋副热带高压面积指数的预报模型。该方法首先将西太平洋副热带面积指数(SI)分解为相对简单的带通分量信号,利用LS-SVM建立各分量信号的独立预报模型,然后对预报结果进行集成。为了评估和比较该方法的预报效果和技术优势,最后比较了在同等条件下WT～LS-SVM模型和神经网络、线性回归模型的独立检验预报效果。试验结果表明,该方法具有泛化能力强、预报精度高、训练速度快、稳定性好、便于建模等优点,具有良好的应用前景。

Based on the method of wavelet decomposition and support vector machine, the area index of western pacific subtropical high forecast model was established. By using this method, the area exponent of western pacific subtropical high was decomposed into several relative simple band-pass signals. Then the independent prediction models of decomposed signals with support vector machine were set up, and independent predicted results were integrated. Finally, in order to assess and compare the effectiveness and technical superiority of the paper used, the independent testing results with different models, such as WT- LS-SVM, neural networks and linear regression model were discussed in the same condition. The testing results showed that the model based on support vector machine exhibited its properties of high forecast accuracy, fast training, high generalization capability and easy modeling.