支持向量机在雷暴预报中的应用

Application of Support Vector Machine to Thunderstorm Forecasting

论文利用2002—2006年AREM模式产品和常规观测报文资料,综合运用改进的K平均聚类和主成分分析等方法,基于MOS原理逐月建立了最小二乘支持向量机和线性规划支持向量机的单站雷暴释用预报模型,并针对海口站2007年58月进行了具体的预报。结果表明:支持向量机结合AREM模式产品进行雷暴的释用预报是合适、有效的,而且主成分分析对预报结果的提高也起到了积极的作用。

In the paper the K-means clustering of the improved algorithm, the principal component analysis （PCA） and other methods are used to establish the interpretation forecasting model of thunderstorm by the least squares support vector machine （LS_SVM） and linear programming support vector machine （LP_ SVM） based on MOS theory monthly in terms of AREM prediction products and conventional observation data during 2002 to 2006. And use the data at Haikou Station for testing from May to August 2007. The results show that, combining with SVM and AREM products to interpret the forecast products is feasible. The PCA also plays a positive role in improving the forecast accuracy.