Characterizing and Learning the Mobile Data Traffic in Cellular Network
Rongpeng Li, Zhifeng Zhao, Qi Chen, Honggang Zhang · 2018
Traffic characterization, learning, and prediction in cellular networks, which is a classical yet still appealing field, yields a significant number of meaningful results. This chapter presents an intensive study on the fundamental traffic nature of mobile instantaneous messaging (MIM) service through a large amount of “Wechat/Weixin” traffic observations from operating cellular networks. It examines the results of fitting the application-level dataset to α-stable models. The chapter explains entropy theory to analyze the feasibility of predicting traffic dynamics theoretically. It demonstrates the microscopic traffic predictability in cellular networks for circuit switching's voice and short message service and packet switching's data service. The chapter aims to fully take advantage of the traffic modeling results. The proposed framework consists of three modules. Among them the “α-Stable Model and Prediction” module would take advantage of the already known traffic knowledge to learn and distill the parameters in α-stable models and provide a coarse prediction result.