A Network Traffic Prediction Model Based on Wavelet Transformation and LSTM Network

Network traffic has apparent characteristic of burst. The time series of it presents nonlinear. It is difficult for traditional linear method to predict it accurately. To solve the problem, this paper proposes to decompose the original traffic to an approximation sequence and several detail sequences with the method of wavelet transformation. On this basis, the change trend of traffic is learned by LSTM network and the burst information is extracted at multiscale to complete the prediction of future traffic. The experimental results show that for the prediction error, the model constructed with LSTM network is superior to the models constructed with LSSVM, BP neural network and Elman neural network. In addition, the model proposed in this paper performs better than the ordinary LSTM network model for predicting the burst of traffic.