An improved echo state network based on variational mode decomposition and bat optimization for Internet traffic forecasting

Internet traffic forecasting plays an important role in network resource allocation, planning and anomaly detection. Developing an accurate and robust Internet traffic forecasting model is very essential. However, Internet traffic usually exhibits nonlinear and multi-scale data character. Thus, Internet traffic forecasting is also a huge challenge. Echo state network (ESN) is a novel time series forecasting approach and it is able to approximate any complex nonlinear relationship. Considering the fact that the standard ESN easily suffers from the influences of initial random weights, this paper employs bat algorithm (BA) to overcome this drawback. In addition, variational mode decomposition (VMD) is utilized to decompose the original Internet traffic series into several band-limited intrinsic mode functions (BLIMFs) for capturing the multi-scale data character. Based on decomposition result, the training set is reconstructed. Verified by four data sets, the proposed model defeats six popular forecasting methods. The simulation results indicate that the proposed model is an effective and robust alternative for Internet traffic forecasting.