A hybrid approach of support vector machine with particle swarm optimization for water quality prediction

Water quality prediction is an important and widely studied topic since it has significant impact on national or regional ecological and water resources management. Due to water quality indicators series nonlinearity and non-stationary, the accuracy of conventional mostly used methods including regression analysis, ARIMA and neural network has been limited. The use of support vector machine has been shown to be an effective technology to solve classification, prediction problem of nonlinearity and small sample. However, the practicability of SVM is effected due to the difficulty of selecting appropriate SVM parameters. This paper presents a hybrid approach of support vector machine with particle swarm optimization to determine svm free parameters for developing the accuracy of predictions. The approach is applied to predict Heishui river water quality of the Beibei, Chongqing. Traditional ARIAM model and BP neural network are investigated as comparison basis. The experiment results show that the proposed approach can achieve better prediction performance.