Network traffic classification using AdaBoost Dynamic
Érico N. de Souza, Stan Matwin, Stênio Fernandes · 2013
Accurate traffic classification and identification is of paramount importance for proper network management and control in both edge and backbone networks. The use of Machine Learning (ML) algorithms has been gaining popularity due to its widespread availability and to its somewhat straightforward application to Internet traffic. This work focus on a specific case of using ML algorithms for network traffic classification. We introduce AdaBoost Dynamic with Logistic Function (AB-DL), an extension of AdaBoost.M1, that combines various classifiers to improve the final hypothesis. We carefully choose parameters from the flow records traces to improve the accuracy of the algorithms. Tests were executed with a publicly available data set from Ground Truth, and the other simulation was executed in a data set generated from University, that is not public. Results show that AB-DL achieve accuracy of 93% and 98.1%, respectively from each data set.