Prediction of Surface Water Supply Sources for the District of Columbia Using Least Squares Support Vector Machines (LS-SVM) Method

Abstract

In this research, we developed a predictive model based on least squares support vector machine (LS-SVM) that forecasts the future streamflow discharge using the past streamflow discharge data. A Gaussian Radial Basis Function (RBF) kernel framework was built on the data set to tune the kernel parameters and regularization constants of the model with respect to the given performance measure. The 10-fold cross-validation is used as a cost function for estimating the performance of the model. The training process of LS-SVM was designed to train the support values and the bias term of an LS-SVM for function approximation. After the network has been well trained, we test the prediction performance on the new testing samples, as well as the training samples. The USGS real-time streamflow data were used as time series input. The experimental results showed that the proposed LS-SVM algorithm is a reliable and efficient method for streamflow prediction, which has an important impact to the water resource management field.