Predictability Analysis Of Mobile User Traffic Usage

Predicting users' traffic usage has great significance for optimizing the cellular network.The performance of cellular network will change significantly as the combination of the network's traffic type changes.Foreknowing the individual's future traffic type,operators can foresee the traffic combination of the specified region of the network,thus can take appropriate measures to optimize the network's performance of the next moment,such as adjusting configuration and deploying resources to the region.Many predicting methods have be applied to predict user's future traffic usage,such as Linear Regression Model(LRM),Auto-Regressive and Moving Average Model(ARMA),Markov chain,Artificial Neural Network(ANN)and Support Vector Machine(SVM).Some location modeling and predicting methods can also be used in predicting traffic usage,such as using historical choices to guide future choices.With the evolution of these prediction methods,prediction accuracy of traffic and application usage is getting higher and higher,but the uncertainty of human behavior still exist,the observer cannot learn of the user's motives and plans,so it is impossible to accurately forecast the traffic usage of the user.This bring us a question:to what extent can we predict user's future traffic usage?Chaoming Song et al.once used entropy theory to explore the predictability of the changes of mobile user position.This paper will use entropy to study the predictability of the mobile phone user's traffic usage based on our real-world data.In this article,we will first describe existing problems in current network architecture,and the status of research of future networks.Secondly,the article will introduces the user behavior research and methodology and prediction algorithms of time-series prediction problem of network traffic usage.Then,we will use entropy theory to analysis the predictability of mobile users,behavior.We will calculate user's behavior entropies based on the massive historical records of user's network usage.After calculating the entropies,we can calculate the predictabilities of the user's behaviors based on the entropies.This theoretical predictability result will guide the innovation of predicting methods,indicating the ultimate degree that a user's future traffic usage can be predicted.Finally,this article also describes the future use of entropy theory to calculate the predictability of user behavior,describes the application of this approach in the future.