A Prediction Model Base on Evolving Neural Network Using Genetic Algorithm Coupled with Simulated Annealing for Water-level

Abstract

In this study, a nonlinear forecasting model is proposed in order to obtain accurate prediction results and ameliorate forecasting performances. In the model, the genetic algorithm (GA) is coupled with simulated annealing (SA) algorithms to evolve a back-propagation neural network (BPNN) algorithm, called GASANN. The new model's performance is compared with three individual forecasting models, namely weighting moving average (WMA), stepwise regression (SR) and autoregressive integrated moving average (ARIMA) models by forecasting yearly water level of Liujiang River, which is a watershed from Guangxi of China. The results show that the new model outperforms than the other models presented in this study in terms of the same evaluation measurements. Therefore the nonlinear model proposed here can be used as an alternative forecasting tool for water level to achieve greater forecasting accuracy and improve prediction quality further.