A grey-neural networks prediction model of death toll in “5.12” Wenchuan Earthquake

Destructive earthquakes often caused huge casualties. In order to reduce casualties, the analysis on the impact factors that determining casualties in the earthquake and the development of rational prediction model to casualties become an important research topic. Because of inaccuracy and ignorance of present prediction method of death toll, a more accurate prediction model is brought up by grey correlation theory and BP neural networks. According to the data collected from 31 hardest-hit counties in “5.12” Wenchuan earthquake, the influencing factors are sorted by grey correlation theory. Furthermore, the data collected from 31 hardest-hit counties are viewed as training samples, and neural networks prediction model is utilized to estimate the death toll of Mianzhu, Mianxian and Wudu counties in Wenchuan Earthquake. Finally, Analysis results of examples show that neural network prediction model, which can approximate complex non-linear problem, can help improve accuracy and reliability of estimation to casualties.