A semi-supervised clustering algorithm for real network traffic with concept drift

Abstract

Traffic classification has been widely applied for networking. Previous works paid much attention to static network traffic. In this paper, we propose a new strategy for the semi-supervised clustering algorithm to deal the concept drift in a dynamic network, as well as updating the model incrementally. Moreover, our algorithm can find new clusters and reduce the impact of noises. The results of simulation demonstrate the effectiveness of semi-supervised clustering algorithm.