Trend analysis of extreme rainfall based on BP neural network

Standing the perspective of data mining and using the basic principles of artificial neural network to establish a average extreme rainfall prediction model which is based on BP neural netwok.This model only use the extreme precipitation indexes as the factors to predict the average extreme rainfall in the coming year.The model combined with stepwise regression to select input vectors and used bayesian regularization method to further improve the generalization ability, thereby increased the forecast accuracy of the trend of extreme rainfall.It is proved that the model is indeed valid and reliable by experimenting on many years´ daily precipitation data of two sites in Yangtze river.