Research on runoff prediction based on wavelet transform and least squares support vector machines model

• Published in: 2014 12th International Conference on Signal Processing (ICSP) pp. 1469 - 1473
• Main Authors: Fanping Zhang, Deshan Tang, Meihong Zhang, Huichao Dai
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2014
• Subjects: Autoregressive processes; Forecasting; least squares support vector machines; Predictive models; runoff predicting; Support vector machines; Time series analysis; wavelet transform; Wavelet transforms
• ISBN: 9781479921881; 1479921882
• ISSN: 2164-5221
• DOI: 10.1109/ICOSP.2014.7015243

A new hybrid model that combines wavelet transform(WT) and least squares support vector machines(LSSVM) called the wavelet least squares support vector machines(WT-LSSVM) model is proposed and applied for runoff time series prediction. Time series of monthly runoff of Tangnaihai Station located in Yellow River upper stream were analyzed by the WT-LSSVM model. The observed time series are decomposed into sub-series using a discrete wavelet transform function and then an appropriate sub-series is used as input to the WT-LSSVM for forecasting hydrologic variables. The hybrid model (WT-LSSVM) was compared with the standard SVM model. The WT-LSSVM model is able to provide a good fit with the observed data. The benchmark results from WT-LSSVM model applications shows that the hybrid model produces better results than the standard SVM model in estimating hydrograph properties.