Water quality prediction based on a novel hybrid model of ARIMA and RBF neural network

Abstract

Improving the accuracy of the water quality prediction is an important and difficult task facing decision makers in water resources management. Many researchers have argued that combining different models can be an effective way of improving upon their predictive performance. The hybrid models of autoregressive integrated moving average (ARIMA) and neural network, as one of the most popular hybrid models for time series forecasting, have recently been shown successfully for water quality prediction. However, these models have many assumptions and limitations. In this paper, a novel hybrid model of ARIMA and Radial Basis Function Neural Network (RBF-NN) is proposed in order to yield more general and higher accuracy prediction model than traditional hybrid ARIMA-ANNs models for water quality prediction. The proposed model consist of an ARIMA model, which was a linear model and used to obtain the existing linear structures, and an RBF-NN model that is used to capture the nonlinear architectures and do the prediction. Experiments results show that the proposed model can be an available and effective way to improve the accuracy of the water quality prediction.