A Concept Drift Detection Approach Based on Jensen-Shannon Divergence for Network Traffic Classification
Wujun Yang, Rui Su, Yuanzheng Cheng, Juan Guo · 2022
Network traffic features change with time and network environment, creating a concept drift problem that leads to a decrease in the accuracy of machine learning-based network traffic classification methods. This is because the traditional network traffic classifiers are static models that cannot adapt to the changes in data distribution. Therefore, we proposed a concept drift detection approach based on Jensen–Shannon divergence, named CDJD. The method uses a double-layer window mechanism to detect changes in data distribution based on the Jensen-Shannon divergence, and thus detects concept drift. After detecting concept drift, the Jensen-Shannon divergence is used to check whether the current concept is a recurrence of the past concept and thus decide whether to reuse the old classifier. The method is experimentally compared with common concept drift detection methods, and the experimental results show that the method can effectively detect concept drift and showing better classification performance.