Network traffic cognition model based on space-time fractals

Considering the problem of traditional fractal (TF) features being difficult to achieve both high accuracy and fast speed in network traffic cognition, the idea of space-time separation was proposed on the basis of fractal theory. With space-time fractal (SF) features generated by the space-time separation, a new traffic cognition system called the space-time fractal model (SFM) was established. In order to obtain SF, the spatial and temporal sequences were observed, and further constructed to generate vectors by Legendre transformation, which were mapped into dual space. The physical significance of SF lied in capturing the characteristics of traffic bursts at different scales of space and time, while TF were the fusion of SF across spatial and temporal scales. Compared with TF, SF represented network traffic more comprehensively and thus were able to identify traffic more accurately. Moreover, SF were more computationally efficient than TF, enabling SFM to achieve high cognition speed as well as strong cognition accuracy. The experimental results show that the cognition performance of SFM is superior to other methods.