Application of Integrated Back-Propagation Network and Self-Organizing Map for Groundwater Level Forecasting

Abstract

In this paper, based on the combination of the back-propagation network (BPN) and the self-organizing map (SOM), a groundwater level forecasting model is proposed, named improved multisite SOM-BPN model. In the proposed model, the SOM is used to determine the number of hidden layer neurons, and the autoregressive integrated moving-average (ARIMA) model and semivariogram are used to determine the number of input neurons. To evaluate the forecast accuracy of the proposed model, the improved multisite SOM-BPN model as well as five other models (ARIMA model, single-site BPN model, single-site SOM-BPN model, multisite BPN model, and multisite SOM-BPN model) are applied to actual groundwater level data in southern Taiwan. According to the results, it is found that the single-site and multisite BPN models can forecast more precisely than the ARIMA model. Moreover, the results show that the multisite model is more competent in forecasting groundwater level as compared to the single-site model. Finally, among the six models, the improved multisite SOM-BPN model has the highest accuracy. The improved multisite SOM-BPN model is recommended as an alternative to groundwater level forecasting because it cannot only produce reasonable forecasts, but also objectively determine the suitable number of hidden layer neurons of BPN.