Mobile Communication Traffic Forecast Based on a New Fuzzy Model

An accurate model and prediction of traffic plays a crucial role in mobile network planning and design. However, it is difficult to obtain an analytical model of the mobile traffic due to the high complexity of the mobile network. In this study, a novel prediction method based on historical traffic data from the mobile networks, which is considered as chaotic time series, is proposed. It is built on the theory of dynamic system reconstruction, the Takagi-Sugeno (TS) fuzzy model and the Support Vector Machines (SVMs). Because those new elements are involved, it can deal with the time series with noise, and has strong robustness. At First, to reconstruct the dynamic system in phase space, the method to calculate a suitable embedding dimension and time delay is discussed according to the mobile traffic time series. Then, the fuzzy model of the dynamic system is set up, and its parameters are obtained by using subtractive cluster and SVMs. Finally, prediction of mobile traffic with the fuzzy model is analyzed and its comparison with TS model is given. The experiment results show that the proposed method can be applied to various chaotic time series with noise.