Network Traffic Prediction Algorithm based on Wavelet Transform

The features of dynamic, noise and instability, make the network traffic eruptive and unstable, and this obstructs the network traffic prediction. In order to figure out its characteristics and developing tendency accurately, the paper proposes a wavelet-transform-based prediction algorithm: Firstly, with the multi-resolution analysis of wavelet transform, the network traffic, which is difficult to be analyzed or modeled in the time domain, is divided into different bands of frequency by wavelet decomposition. Later, simplify and stabilize the divided traffic by denoising with different thresholds on the detail components of different sub-bands of frequency. At last, synthesizes the traffic prediction with the predictions of the denoised sub-traffic by Auto Regressive Moving Average Model. The algorithm improves the prediction accuracy significantly in practical modeling, especially in short-range traffic prediction.