A Research Of Sampling Technology Applied In Complex Network

With a wide range of applications in recent years, the study of the complex network attracts high attention of lots of scholars in different fields. Most of them have intensely worked on the topological property and mathematical model of the real networks. However, the scale of the real networks in fact are quite large, we are unable to acquire the whole information due to the limit of the technology and cost. Because of that, a number of the empirical studies on complex network are based on the incomplete network data. Therefore, it is a big question of the complex network study, which is how to sample reasonably in the network to estimate the property of the complete network. The design of sampling methods and the analysis of sampling effect are very meaningful to the network under the big data background. In this paper, the author uses four different sampling methods to sample in three types of complex networks, and calculates the network topological properties of the sample networks. By c the author concludes that the different sampling methods corresponding to the estimation of different topological properties.The first part(para.1) is the research background. The author introduces the importance and necessity of complex network sampling under the current background. At the same time, The author also analyzes the studies and implementations of complex network sampling both abroad and in China, and expounds the purpose and significance of this research.The second part(para.2) is the theoretical analysis. Firstly, the author introduces the theoretical basis of complex network sampling, including its historical developing process, the characteristic and construction method of the classical complex network models, and some common network topological properties. Secondly, the author describes the main idea of three classical sampling methods. At the end, a new sampling method, the 2nd stage sampling method is put forward.The third part(para.3-5) is methods application. The author takes samples from the BA scale-free network, WS small-world network, EU email communication network respectively with the completely random sampling, random walk sampling, snowball sampling and the 2nd stage sampling methods in the different sampling rate. By comparing the estimated effect of topological properties, the author concludes the features of different methods, and the strengths and weaknesses when sampling in different networks.The last part(para.6) is summary. The author lists different sampling method fitted with different topological estimation of complex networks. At the end of this paper, the innovation points and shortcoming of the research are pointed out.