The Evolution Analysis And Prediction Of Connection Mechanism In Internet Topology

Internet is a classical instance of complex network. The analysis and modeling on its topology has become a hot topic at present. Recently, related research focus on the analysis of statistic characteristic concerning topology. However, most of the internet evolve models currently studied are phenomenon-based models, which may not be able to represent the essence of internet evolution. Evolution mechanism proposed in some topology models, such as preferential attachment and internal link growth, need the further analysis and validate in the true internet topology's data.In this dissertation, the research based on massive data authorized by CAIDA(The Cooperative Association for Internet Data Analysis)Skitter project. The topology's evolvement of Internet autonomous system level was analyzed from time series viewpoint and the data's time range is from January 2000 to September 2006.First of all, the time evolution of basic statistic was analyzed. Then based on a new statistical method, the behaviors of preferential attachment mechanism were also analyzed in actual Internet. The result of analysis shows that the preferential attachment mechanism itself shows a variation over time in the actual internet evolution. The predominance of the large degree nodes to get connection with the new nodes weakens over time;Then by analyzing the internal link growth of Internet topology, the interactive-growth mechanism described in PFP models is inexistent in the actual internet evolve.The new internal links is actually around the nodes with small degree.On the contrary, the links between large degree nodes shows a trend of decrease. Analysis also shows that the rewire mechanism of internal link is also an important impetus to the evolution of network topology.Finally, in order to further predict the evolution development of Internet topology, the classification method by degree was introduced. And use this way to depict the preferential attachment characteristic in Internet's growth.For eliminating the deficiency in data set, nodes were divided into weak nodes and strong nodes.By analyzing the variation law of the weak nodes'preferential probability, GA(Genetic programming Algorithm)was used to get the most precise prediction model by curve fitting. Experiment shows that the prediction model can predict the preferential probability within a certain range, and can re-produce the evolution law of preferential attachment mechanism in actual Internet topology. The model can make a qualitative evaluation for Internet's preferential attachment mechanism in future years.The evolution analysis for preferential mechanism provide a new idea for research and ameliorating the current network topology model and design next-generation Internet routing protocols.