Method of periodic dynamic pattern mining based on complex network

Traditionally, analysis of dynamic network has been focused only on a single snapshot or integrated network obtained over a period of time. However, the temporal feature in dynamic network has been ignored. In this paper, periodic dynamic pattern (PDP) mining method is proposed to discover the dynamic evolution rule of social networks. We define a periodic dynamic pattern based on complex network, and construct matrix of graph (MG) to present the evolution of complex network over time. Moreover, a regular edge pattern searching (REPS) algorithm based on MG is designed. It is used to select frequent and regular edge existence sequence (EES) in MG with progressive scan. To improve efficiency of the algorithm, sequences according with frequency threshold and REP should be transformed to decimal integer. Then, depth-first search method is used to mine PDP, which owns the same REP in dynamic network. So, we can predict future network evolution and control its behavior. Experimental results demonstrate that our method is an effective periodic dynamic pattern mining. Copyright © 2015 Binary Information Press.