Neural Network Application Based on GIS and Matlab to Evaluation of Flood Risk

Abstract—In order to test the Artificial Neural Networks (ANN) in the applicability of flood risk assessment, this paper applies the traditional BP neural networks (BPNN), radial basis function neural networks (RBFNN) and probabilistic neural networks (PNN) to establish flood risk assessment model using MATLAB combined with GIS technology. It is observed that BPNN is superior among three methods. Taking Beijiang River basin as a case study, the risk assessment map based on BPNN model shows that the dangerous areas are mainly located in these areas: Sihui, Qingyuan city, Fogang, northwest Huaiji, central Yangshan, central Yingde, northeast Nanxiong and so on. Compared with a few historical large floods, above results can better reflect the actual situation of flood risk in Beijiang River basin, which validate the rationality of the presented model and provide a reference for flood control and disaster assessment. Keywords-flood damage; risk assessment; BP neural network; geographic information system I