Traffic classification using cost based decision tree

A novel method for achieving practical real-time traffic classification is proposed in this paper, which is based on C4.5 decision tree. Most existing traffic classification algorithms only focus on accuracy of the classification results, but lack of considering the various costs in actual deployment. So they cannot guarantee that the obtained tree construction is optimal for hardware and software processing. To solve this problem, our Cost Based Feature Evaluation procedure defines UnitGainRatio as the metric of attributes to find the best tree construction when considering the attribute acquisition and processing cost. We also introduce another method called Fuzzy Delicacy Node Selection procedure to choose the more suitable node, when their UnitGainRatio are too close to each other. The experiment results show that the proposed method reduces the average cost compared with similar algorithm.