基于支持向量机的海南气温预测模型研究

Temperature Forecast Model Based on Support Vector Machine Method

依据CMSVM2.0函数估计和交叉验证1等方法,利用1970~2014年海南省各市县月平均气温资料,建立气温的SVM回归方法预报模型,并进行了模拟实验.实验结果表明支持向量机算法在气温短期预测中具有良好的预报能力,其中交叉验证1的预测效果略高于业务预报,尤其在冬季、夏季和秋季均有较好的预报能力.此外,SVM对海南省北部、西部、南部市县的预报效果较好,而对中东部市县的预报效果相对较差.

Based on the ＂ Function Estimation＂ and ＂ Cross Validation 1＂ of the CMSVM2. 0,the temperature data of Hainan from 1970 to 2014 were used to construct the forecasting model of regression method of SVM,and the simulation experiments were performed. The results indicated that the CMSVM2. 0 has good forecasting ability for short-term temperature forecast of Hainan,and the prediction effects of the ＂ Cross Validation 1＂ is higher than that of the general prediction,especially,in the winter,summer and autumn. Additionally,the prediction effects of SVM for the west,north and south of Hainan are better than that for the east and central.