Traffic profiling in mobile networks using machine learning techniques

Abstract

This paper tackles a problem of identifying characteristic usage profiles in the traffic related to packet services (PS) in mobile access networks. We demonstrate how this can be done through clustering of vast amounts of network monitoring data, and show how discovered clusters can be used to mathematically model the PS traffic. We also demonstrate accuracy of the models obtained using this methodology. This problem is important for accurate dimensioning of the infrastructure for mobile access network. © 2011 Springer-Verlag.