Mobile network providers face an ever-increasing number of mobile devices requesting similarly increasing amounts of data. In this article, we present a two-step approach to modeling and simulating the amounts of data produced by mobile devices based on applications that are highly utilized on the network. In the first step, we separate the applications on a mobile device into highly utilized and background ones for the overall population to be modeled. With the identified overall application groups, we employ a four-state Hidden Markov Model to capture the characteristics of the high utilization applications as aggregates per device; the characteristics of the background applications are matched to four states, dependent on the high utilization aggregates’ states. Utilizing the Exponential distribution for both, we closely match their original user-based characteristics. The suitability of our model is lastly corroborated through simulation-based comparisons of estimations for the bandwidth requirements of the individual users; or model’s estimates are typically within ten percent of the original values.
|State||Published - Feb 2015|