A method of combining traffic classification and traffic prediction based on machine learning in wireless networks

TL;DR

This paper presents a machine learning-based method that jointly performs traffic classification and prediction in wireless networks, improving accuracy by feedback between classifiers and predictors.

Contribution

It introduces a combined approach where classification and prediction inform each other, enhancing performance over traditional independent methods.

Findings

Improved accuracy in traffic classification.

Enhanced traffic prediction precision.

Effective feedback mechanism between classifiers and predictors.

Abstract

With the increasing number of service types of wireless network and the increasingly obvious differentiation of quality of service (QoS) requirements, the traffic flow classification and traffic prediction technology are of great significance for wireless network to provide differentiated QoS guarantee. At present, the machine learning methods attract widespread attentions in the tuples based traffic flow classification as well as the time series based traffic flow prediction. However, most of the existing studies divide the traffic flow classification and traffic prediction into two independent processes, which leads to inaccurate classification and prediction results. Therefore, this paper proposes a method of joint wireless network traffic classification and traffic prediction based on machine learning. First, building different predictors based on traffic categories, so that different types of classified traffic can use more appropriate predictors for traffic prediction according to their categories. Secondly, the prediction results of different types of predictors, as a posteriori feature, are fed back to the classifiers as input features to improve the accuracy of the classifiers. The experimental results show that the proposed method has improves both the accuracy of traffic classification and traffic prediction in wireless networks.