Leakage forecasting for water supply network based on GA-SVM model

Support Vector Machine (SVM) had excellent learning, regression ability and generalization ability, which used structural risk minimization based on smaller sample. The parameters of SVM have important effect to its performance. For this reason the parameters selection is the very important research content of the SVM. The traditional parameter selection method is time-consuming and difficult to obtain the optimal parameters because it is based on artificial tests. To this problem, one kind of method to choose the parameters of the SVM by genetic algorithm(GA) is proposed in this paper. The parameters of SVM model are pretreated through genetic algorithms to get the optimum parameter values, and these parameter values are used in the SVM model and genetic algorithm-support vector machine (GA-SVM) model is obtained, which will be used to make leakage forecasting for water supply network. The experiment result shows the SVM regression model optimized by GA have high forecast accuracy, and GA is one kind of effective method for SVM parameters choosing.