LNTP: An End-to-End Online Prediction Model for Network Traffic
Lianming Zhang, Huan Zhang, Qian Tang, Pingping Dong, Zhen Zhao, Yehua Wei, Jing Mei, Kaiping Xue
Abstract
As network data keeps getting bigger, deep learning is coming to play a key role in network design and management. Meanwhile, accurate network traffic prediction is of critical importance for network management that is implemented to improve the quality of service (QoS) for users. However, the performance of existing network traffic prediction methods is still poor due to three challenges: complicated characteristics of network traffic, dynamics of traffic patterns caused by different network applications, and a complex set of variations like burstiness. In this article, we propose a long short-term memory (LSTM) based network traffic prediction (LNTP) model, which aims to forecast network traffic timely and accurately. The model can be divided into two parts, namely, wavelet transform and LSTM. The working process of LNTP falls into three stages, i.e., data acquisition, model training, and online learning and prediction. In addition, to avoid the negative incentives to models caused by the burstiness and adapt to the changing trend of the network traffic, a weight optimization algorithm of the neural network named sliding window gradient descent (SWGD), is also proposed. Extensive experiments based on two real-world network traffic datasets demonstrate that our model outperforms the state-of-the-art network traffic prediction models by more than 29 percent.