Flood Disaster Evaluation Based on Improved BP Neural Network Model

Floods have become increasingly alarming worldwide. In this paper, we firstly introduce the Improved BP Neural Network model (IBPNNM) based on entropy value theory in detail. Then we adopt Levenberg-Marquardt (LM) algorithm to achieve a higher speed and a lower error rate to overcome the shortcomings of the traditional BP algorithm as being slow to converge and easy to reach extreme minimum value. To illustrate the procedure of the proposed method, we apply IBPNNM to describe the flood disaster risk quantitatively in China by using statistical data respectively, including number of deaths, number of victims, collapse of housing, affected crop area, direct economic loss of 22 Areas in China in 1998. The results of IBPNNM are compared with the results of the matter-element analysis method to confirm that the proposed model is reasonable, effective and feasible.