Study on Prediction Model for Buffer Memory Based on SVM
Yongfa Ling, Yali Gao · 2010
The key to high performance network design is the ability to build model and make prediction for performance parameters. Configuration of buffer memory directly influences the delay and loss rate of network. Good match between buffer memory of network and network capacity will improve performance of network. Therefore, in this paper, Support Vector Machine (SVM) is used to predict the business flow data of network, and sample is trained for distribution rules of data beyond sample. Then, prediction model for queue buffer memory of network is designed. It is suggested by experimental data that the model is of high training efficiency and prediction accuracy.