Application of Chaotic Time Series and Wavelet De-noising to SVM Model and Its Application
Ziai Lu · Water Resources and Power · 2010
Base on the wavelet de-nosing theory,chaotic characteristics of the monthly runoff series in Beibei hydrologic station is analyzed and an improved small data method is applied to calculate the largest Lyapunov index.The phase space is reconstructed by C-C method so that the information of runoff is investigated.Penalty factors and kernel width are optimized with SCE-UA algorithm and RBF kernel function is used to simplify solution procedure of non-liner problem.The results show that the unification of precision and practicability is realized by the SVM runoff prediction model and it can deal with complex hydrologic time series effectively with good generalization ability and prediction accuracy,which provides a new runoff forecast method for lack of data watershed.