Abstract—With the increasing popularity of mobile devices and applications, emerging mobile network traffic exhibits special characteristics in temporal scale e.g., there is a scale variance between the network traffic on weekdays and on weekends. Although most existing methods have been applied to data traffic prediction, few of them take such characteristic into consideration. In this paper, by using real data in mobile networks, we adopt the entropy theory to reveal that the duration of time-series given for prediction doesn't always have a positive impact and that the uncorrelated preceding time-series also deteriorates the prediction accuracy. In view of this, partitioning the network traffic prediction into weekdays’ and weekends’ perspective, we propose a method to predict the data traffic. Finally, we evaluate the proposed method through predicting the data traffic for a future time according to the historical data traffic in a real mobile network. In comparison with the work based on ARMA (Auto Regressive Moving Average) method, our proposed method can reduce the Mean Absolute Percentage Error (MAPE) by 35.7% and 43.8% on weekdays’ and weekends’ prediction, respectively.

Index Terms—Mobile networks, data traffic prediction, entropy theory, time-series, real data

Cite: Yunjian Jia, Beili Wan, Liang Liang, Qian Zhao, Yu Zhang, and Liang Tan, “A New Method for Traffic Prediction in Emerging Mobile Networks," Journal of Communications, vol. 10, no. 12, pp. 947-954, 2015. Doi: 10.12720/jcm.10.12.947-954