원문정보
초록
영어
Network traffic prediction is one of the significant issues. The model for network traffic prediction should meet the following requirements. First, the model should be taken into consideration the characteristics of the network flow such as burstiness, long-range dependence, periodicity and self-similarity. To achieve this, we decompose the original flow in a multi-scale manner into a set of linear and stable representations, and introduce chaos theory to improve the diversity and search coverage. Second, the model should be efficient and accurate. To this end, we propose a prediction model based on SVR, and utilize Ant Colony Optimization (ACO) algorithm for parameter selection of SVR. Besides, we conduct experiments to evaluate the proposed model.
목차
1. Introduction
2. Related Work
3. Network Traffic Prediction Model
3.1. Empirical Mode Decomposition (EMD)
3.2. Support Vector Regression (SVR)
3.3. SVR Improved by Chaos Theory and ACO (SVR-CACO)
4. Experiment
5. Conclusion
Acknowledgements
References