Network Traffic Forecasting Using IFA-LSTM
A network traffic prediction model is built based on LSTM neural network optimized with the improved firefly algorithm (IFA-LSTM). Aiming at some disadvantages of firefly algorithm including slow convergence, and easy to fall in local optimal values, we introduce a location update strategy based on the diversity of population, to avoid the optimization to fall into local optimal values. A dynamic step length updating measure is proposed to improve accuracy of the optimization, and to avoid the optimal solutions’ oscillation problem. Simulation examples show that the prediction accuracy and convergent speed of the IFA-LSTM method are obviously improved,it can be used to predict network traffic.
KeywordsNetwork traffic Firefly algorithm LSTM
This work was supported by State Grid Technology Project ‘Edge Computing Research in Smart Grid Application and Security’ (Grant. 52110118001H); National Natural Science Foundation of China (Grant. 61702048).
- 1.Qiu, J., Xia, J.B., Wu, J.X.: Research and development of network traffic prediction model. Comput. Eng. Des. 33(3), 865–869 (2012)Google Scholar
- 3.Wang, J.: A process level network traffic prediction algorithm based on ARIMA model in smart substation. In: 2013 IEEE International Conference on Signal Processing, Communication and Computing, pp. 1–5. IEEE (2013)Google Scholar
- 5.Dorigo, M., Birattari, M.: Ant colony optimization. In: Encyclopedia of Machine Learning, pp. 36–39. Springer, Boston (2011)Google Scholar
- 7.Malhotra, P., Vig, L., Shroff, G., Agarwal, P.: Long short term memory networks for anomaly detection in time series. In: Proceedings of Presses universitaires de Louvain, p. 89 (2015)Google Scholar
- 9.Medsker, L.R., Jain, L.C.: Recurrent Neural Networks. Design and Applications 5. CRC Press, Boca Raton (2001)Google Scholar