Runoff simulation Based on Least Square Support Vector Machine
- DOI
- 10.2991/iccahe-16.2016.138How to use a DOI?
- Keywords
- Statistical Learning Theory; Least Square Support Vector Machine; Runoff; Simulation
- Abstract
River runoff is highly nonlinear as affected by the combination of climate, underlying surface condition, etc. Prediction of runoff may guide engineering design, construction and operation. Statistical Learning Theory (SLT) studies the rules of machine learning with finite samples. Support Vector Machine (SVM) is a new machine learning method based on Statistical Learning Theory. It is a solution for the highly nonlinear classification and regression in sample space. Map the one-dimensional runoff series input space of one hydrologic station of the Yellow River onto high-dimensional input space. Then calculate the embedding dimension of runoff time series and reconstruct runoff series into three-dimensional phase space. Using radial base kernel function to learn from 83 training samples through grid search method and optimize model parameters to estab-lish the Least Square Support Vector Machine (LSSVM) prediction model of river runoff. Fitting mean-square error of the model is 0.0148. Prediction mean-square error of the model on 20 samples is 0.0120, a correlation coefficient of 0.975 between predicted and measured values. The result shows that the generalization ability of LSSVM model is high and the prediction result is satisfacto-ry..
- Copyright
- © 2016, the Authors. Published by Atlantis Press.
- Open Access
- This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).
Cite this article
TY - CONF AU - Jun Ping Liu AU - Jun Jie Zhou AU - Xian Bai Zou PY - 2016/10 DA - 2016/10 TI - Runoff simulation Based on Least Square Support Vector Machine BT - Proceedings of the 2016 5th International Conference on Civil, Architectural and Hydraulic Engineering (ICCAHE 2016) PB - Atlantis Press SP - 885 EP - 890 SN - 2352-5401 UR - https://doi.org/10.2991/iccahe-16.2016.138 DO - 10.2991/iccahe-16.2016.138 ID - Liu2016/10 ER -