A Trend Filtering Based Prediction Model on Network Traffic
Currently, the most effective traffic prediction methods are based on ANNs (Artificial Neural Networks), for their great ability to process complex datasets. However, ANNs-based prediction models require large training datasets to identify parameters and optimize structures. Furthermore, more time on training limits their real-time applications. In this paper, we develop a modification of Hodrick-Prescott (HP) trend filtering model with the constraints of local structure and periodical property of network traffic. We preserve the locality by smoothing the neighboring traffic measurements, and maintain the periodicity by keeping the similarity of daily and weekly traffic series. Then the generated trend intensifies the characteristics of network traffic and is easier to be learned. Moreover, we predict traffic volumes through the prediction of trend such that both the number of training samples and the corresponding training time being reduced. We process two datasets in our experiments. The results show that our method performs better than BP (Back Propagation)-based prediction algorithm by 24.5% in MAE (Mean Absolute prediction Error).
KeywordsTraffic prediction Trend extraction Behavior patterns
This work was supported by the Scientific Research Project of ChongQing Board of Education Committee (KJQN201801103), and the Doctoral Program of Chongqing Federation of Social Science (2018BS68).
- 10.Groschwitz, N., Polyzos, G.: A time-series model of long-term traffic on the NSFNET backbone. In: IEEE International Conference on Communications (1994)Google Scholar
- 11.Wang, L.P., Teo, K.K., Lin, Z.P.: Predicting time series with wavelet packet neural networks. In: IEEE International Joint Conference on Neural Networks, pp. 1593–1597 (2001)Google Scholar
- 13.Lu, Z., Guangwei, B.: Multifractal model based on gamma distribution for network traffic. Comput. Eng. Appl. 47(19), 124–127 (2011)Google Scholar
- 14.Vieira, F.H.T., Bianchi, G.R., Lee, L.L.: A network traffic prediction approach based on multifractal modeling. J. High Speed Netw. 17(2), 83–96 (2010)Google Scholar
- 15.Li, Z., Lei, Q., Kouying, X., Xinyan, Z.: A novel BP neural network model for traffic prediction of next generation network. In: International Conference on Natural Computation (2009)Google Scholar
- 19.Zhu, M., Wang, L.P.: Intelligent trading using support vector regression and multilayer perceptrons optimized with genetic algorithms. In: International Joint Conference on Neural Networks (2010)Google Scholar