Analysis and practicality of traffic flow classification method using binned time-series data

Abstract

It is important for network operators to know how users use the network to plan and design network facilities. In this paper, we propose flow classification method using binned time-series data in consideration of practicality. In this paper, we evaluate the proposed method by using mobile data traffic and show the method can maintain classification accuracy as compared with method using the whole flow information. We also confirm applicability to classification of encrypted flow.