The traditional prediction methods are hard to describe the evolution rule of runoff which is a complicated non-linear time sequence system.Based on 1957-2003 annual runoff observed data in Bajiahu hydrological station in the Jingouhe River Basin,a runoff prediction model with RBF neural network was established in Matlab environment after the subtraction of the data was studied using normalization and autocorrelation algorithms.The results indicated that the prediction model with RBF neural network had high...