2016, 38(9): 2180-2187.
doi: 10.11999/JEIT151338
Abstract:
To study the effects of information coupling of local topology on the complex network evolution, a new weighted method is proposed based on local topology information, which can measure the closeness of connection and the coupling degree of topology information between nodes. In this paper, to demonstrate the efficiency of the information coupling of local topology, an empirical research is made on characteristic statistics of evolving model and real network data testing of link prediction respectively. Firstly, the weighted method is applied to BA model; TwBA and the local world model TwLW are proposed based on the topology weighted method. Simulation experiments show that the degree distribution of TwBA can be rapidly changed from exponential distribution to power law distribution with the increasing of the connection numbers for new added nodes, which confirmes that the phenomenon of accelerating growth appears widely in the evolution of many real scale-free networks. Then, based on TwBA model, an accelerating growth model A-TwBA is proposed, and the A-TwBA model presents power law distribution for different accelerating growth rates. The degree distribution of TwLW is changed from stretched exponential distribution to power law distribution for different sizes of local world. Finally, the proposed weighted method is applied to link prediction methods (including CN, Salton and RA index), and three weighted indices are proposed. Empirical study shows that the weighted proposed method can significantly improve the prediction accuracy of these basic indices, and some of them are higher than those of the global indices.