Water level prediction based on improved grey RBF neural network model

For in RBF neural network prediction results by random sample of thus affecting prediction accuracy, using the grey prediction model of RBF network is trained, can weaken the randomness of data greatly, so the combination of neural network and grey prediction, construct grey RBF neural network by network model, and hydrological forecasting can improve the accuracy of hydrological forecast. But if the gray scale data is large, due to the parameters of the model of GM (1,1, θ), leads to poor prediction accuracy. In this regard, GM (1, 1, θ) model and use ant colony algorithm to improve its, and the prediction precision can be improved In the construction of RBF network, due to the implicit function node has been relying on the actual experience to determine, with instability, and choose to use the golden section method to determine the hidden nodes. The forecast results show that the grey RBF neural network forecasting model has higher precision and better generalization ability, and it has practical value.