Because network traffic is complex and the existing prediction models have various limitations, a new network traffic prediction model based on wavelet transform and optimized support vector machine(ChOSVM) is proposed. Firstly, the network traffic is decomposed to the scale coefficients and wavelet coefficients by non-decimated wavelet transform based on suitable wavelet base and decomposition level. Then they are sent individually into different SVM with suitable kernel function for prediction. The parameters of SVM are selected by chaos particle swarm optimization. Finally predictions are combined into the final result by wavelet reconstruction. Experiments on network traffic of different time granularity show that compared with other network traffic prediction models, ChOSVM has better performance.