Mobile device-level data modeling through high utilization mobile applications

In this paper, we present a mobile-device level approach to estimating the network data (traffic) that is generated over time. While efforts oftentimes utilize complex approaches, our model captures the main characteristics in the time and data domains of a high utilization application class as Hidden Markov Model while modeling the remaining applications' characteristics in form of a simple background process. We find that our approach is capable of matching the average amounts of data behavior of the source dataset (with a reduction in overall variability of the simulated produced traffic as drawback) and is thus suitable for high level capacity evaluations.