A novel approach to extreme rainfall prediction based on data mining

A novel extreme rainfall prediction model combined with data mining is proposed in this paper. Because of the special nature of hydrological data, our model uses the clustering method to group the next year's average extreme rainfall and then establish a hybrid extreme rainfall prediction model based on building neural networks for each group. Furthermore discriminant analysis is employed to classify the sample's coming year's average extreme rainfall and corresponding BP neural network is chosen to obtain the prediction. We also take the impact of discrimination error into account in our computation of the average extreme precipitation predictive value owing to the discrimination error. Experimental results validate our method and the prediction accuracy is satisfactory.