A Risk Assessment Model of Flood Based on Information Diffusion Method and BP Neural Network

Climate change has caused more frequent floods in China which have already resulted in huge losses. Thus flood risk assessment and management is an important research topic. In this paper, a new model of flood risk assessment is proposed based on the information diffusion theory and the back propagation (BP) neural network. Due to the fact that flood statistics data are relatively short and often insufficient for flood risk assessment, the information diffusion method can transform imperfect flood historical data from a point in a traditional data sample to a fuzzy data set and obtain optimized data sample. Then, the optimized data are used to train neural networks with back propagation and can improve neural network adaptive ability. The flood data of Dongting Lake’s different encirclement dikes are used to assess the flood risk of Dongting Lake with the proposed model in this research. The results are consistent with the actual situation of Dongting Lake area, which thus verifies the model’s effectiveness for flood risk management. This method can be easily applied to effectively resolve problems of insufficient samples in flood risk assessment. Streszczenie. W artykule zaprezentowano nowy model oceny ryzyka powodzi bazujący na teorii dyfuzji informacji I wykorzystujący sieci neuronowe. Dane statystyczne o powodziach są relatywnie krótkie i często niewystarczające do oceny ryzyka. W pierwszym etapie przetwarza się dane historyczne do otrzymania bardziej kompletnych danych. Te dane wykorzystane są do trenowania sieci neuronowych. (Model oceny ryzyka powodzi bazujący na metodzie dyfuzji informacji i wykorzystujący sieci neuronowe)