Application Of Svm Methods To Forecasting Peak Water Level In Tidal River

Water level series in tidal rivers show the complex non-linear characteristics of dependency, mutation and random. Due to the complex interaction mechanism between the flood and the tide, there is a limited awareness about it for people,so forecasting water level in tidal rivers has been very complex and difficult.Support vector machine (SVM) is a new universal learning method from the 1990s based on the VC dimension theory and structural risk minimization principle advanced by Vapnik and others, and it is made to deal with nonlinear classification and nonlinear regression problems and to improve the weaknesses of traditional neural network theory.SVM works well in solving the highly nonlinear classification and regression problems of the sample space, and has been successfully used in classification, function approximation and time series prediction and so on.This paper analyzes the current research progress of flood forecast methods in tidal rivers and the application status of support vector machine, discusses the feasibility of applying SVM in flood forecast in tidal rivers,and outlines the principle of SVM methods and the LIBSVM software. The paper selects predicting factors and constructs sample data sets by analyzing the cause and influencing factors of the floods in Puyangjiang River, then selects RBF as the kernel function by testing, and prefers model parameters by using cross-validation and grid search methods, finally builds SVM prediction model of peak water level in Meichi station.