Abstract: This paper proposes a combination network traffic prediction model based on neural networks.Firstly, SMOF network is used to cluster the network traffic data, and then used the clustering Elman network to train and predict, and Elman network is optimized by genetic algorithm to improve the network traffic prediction accuracy.Simulation results indicate that the combined neural network prediction model accelerates the prediction speed and improve the network traffic prediction accuracy, overcoming the single forecasting model insufficient, it has the very good application prospects.