Application of least square support vector machine for thunderstorm prediction

Abstract

SVM possess great potential and superior performance owing to the structural risk minimization (SRM) principle in SVM that has greater generalization ability and is superior to the empirical risk minimization (ERM) principle as adopted in neural networks. Considering the characteristics of the thunderstorm in Chongqing, the thunderstorm prediction model based on least square support vector machine (LS-SVM) is established. The data are preprocessed and analyzed. Then the samples affecting the generation of thunderstorm in Chongqing are selected, and the modeling process and parameters selection are analyzed. Lastly, Comparing with neural network and standard SVM, the results show that the LS-SVM model has better prediction results and can meet the requirement of practical prediction. The thunderstorm prediction system of Chongqing area has been developed based on the LS-SVM model.