User behavior analysis based on segmentation, clustering and timing relationship analysis.

Analyzing the user behavior is of great importance to the management and control of network users. User behavior, which is actually s set of data exchanges, is ultimately displayed as traffic flow, showing regularity along time. This paper explored this regularity through studying the timing relationship between traffic flows, and proposed a novel user behavior analysis method. The proposed method utilized three steps, which were based on fractal model, improved maximum distance clustering method, and Apriori algorithm for segmentation, clustering, and timing analysis, respectively, to obtain the user behavior regularity from user data exchange. Theoretic analysis and simulation results illustrate that this method can satisfactorily classify the packet sequence and discover the regularity of user transmit behavior without decrypting user information. [ABSTRACT FROM AUTHOR]