HomeUE Research Bulletinvol. 13 no. 1 (2011)

Application of Artificial Neural Network Multi-Layered Perceptron in Flood Forecasting Implemented in C++

Ronaldo D. Milagroso

Discipline: Earth Science



The study developed an artificial neural network (ANN) model using the Multiple Layer Perceptron (MLP) to forecast estimates of flood depth in centimeters. The ANN MLP model was implemented in C++ programming. The multilayered perceptron was selected based on its hyperbolic tangent capability to handle the scaling requirements of the study. The resulting flood forecasting model had an average validation error (1.029) and a greater coverage of the variability (R square = 0.965) of the forecasted output from the sample data, and later determined the importance of the following independent variables: temperature (0.354), humidity (0.321) and flood index (0.180).