Network Traffic Prediction Based on a CNN-LSTM with Attention Mechanism

Abstract

With the increase of Internet users and the rapid development of various large traffic network applications such as video and games, the network traffic load increases sharply, and network congestion occurs from time to time. Existing routers cannot solve this problem. Network expansion is a costly solution, so intelligent routers should be adopted. Traffic prediction is the precondition of intelligent routing algorithms. Because network traffic has strong nonlinear characteristics, the traditional linear model has great error in predicting network traffic. Machine learning algorithms such as support vector regression (SVR), long-term and short-term memory network (LSTM) and convolutional neural network long-term and short-term memory network (CNN- LSTM) have achieved certain results, but their prediction accuracy still can not meet the requirements of practical application. In this paper, a CNN- LSTM with attention mechanism is used to efficiently extract timing information and other relevant feature information, so as to achieve a more accurate prediction of network traffic. The experiment of this paper used the network traffic data of China Telecom in Guangdong Province as an example to predict. The results show that the performance of the model is satisfactory.