Prediction of COD Time Series Based on Wavelet and LS-SVM

MBR(Membrane Bio-Reactor)' s COD(chemical oxygen demand) have important effects on entire system.We design our wavelet transformation and LSSVM(Least Squares Support Vector Machine) method based on COD's nonlinear characteristic.First,we use Mallat deconstruction and reconstruction arithmetic to get trend and random time series,and then we use LSSVM to predict separately.Finally we combine every time series to get final result.Through simulation,our method show higher precision.