Abstract: Support Vector Machine(SVM)is an intellectual learning method based on the statistics theory. The SVM can solve problems of complicated nonlinear pattern recognition of spatial samples. Drought is a respond of water deficit that resulted from the complicated nonlinearity interrelationship of climate factors. Examined fifteen climate factors (southern oscillation index, subtropical high strength index and polar vortex strength index, etc.) by using the SVM method, this study had developed Zhejiang autumn drought prediction model which was based on the RBF kernel function of SVM. The best parameters of SVM for Zhejiang were determined by applying the cross validation method. The autumn droughts predicted by the model of this study agreed well with the truth facts. These results demonstrated the autumn drought prediction model with a better accuracy rate could act as an effective approach of switching climate factor prediction to meteorological hazard prediction.

Key words: font-family: 宋体, mso-bidi-font-family: 宋体, mso-font-kerning: 1.0pt, mso-ansi-language: EN-US, mso-fareast-language: ZH-CN, mso-bidi-language: AR-SA">Support Vector Machine(SVM), Pattern recognition, Drought prediction