Author(s): Sungwon Kim

Linked Author(s):

Keywords: EDRNNM; Reliability Analysis; Uncertainty; Cross-Validation; Flood Stage Forecasting

Abstract: Elman Discrete Recurrent Neural Networks Model (EDRNNM), one of special type neural networks model, is developed to be a highly suitable flood stage forecasting tool yielding a very high degree of flood stage forecasting accuracy at Musung station of Wi-stream, one of IHP representative basins in South Korea. A relative new approach, EDRNNM, has recurrent feedback nodes and virtual small memory on the networks structure. 135 different training patterns, which involve hidden node, standardization method, data length and lead hour, were selected to minimize the structural uncertainty during training performance. And, a cross- validation method is applied to reduce the overfitting problem and select the best one of the 6 training patterns during validation performance. The model parameters, optimal connection weights and biases, are estimated during training performance and they were applied to evaluate model validation performance. From model training and validation performance, EDRNNM is proved to be an outstanding model for flood stage forecasting in the small catchment, Wi-stream basin. However, even if much of unknown structural uncertainty is eliminated during training and validation performance, some of uncertainty remains on input data information. Therefore, the continuous research is required to reduce useless input data during investigation of the relative importance of each of the input data for reliability analysis of flood stage forecasting.

DOI:

Year: 2005