Ensemble Feature Selection and Classification of Internet Traffic using XGBoost Classifier

Abstract

Identification and classification of internet traffic is most important in network management to ensure Quality of Service (QoS). However, existing machine learning models tend to produce unsatisfactory results when applied with imbalanced datasets involving multiple classes. There are two reasons for this: the models have a bias towards classes which have more samples and they also tend to predict only the majority class data as features of the minority class are often treated as noise and therefore ignored. Thus, there is a high probability of misclassification of the minority class compared with the majority class. Therefore, in this paper, we are proposing an ensemble feature selection based on the tree approach and ensemble classification model using XGboost to enhance the performance of classification. The proposed model achieves better classification accuracy compared to other tree based classifiers.