Key Problem Researching In User Sampling, Measurement And Evaluation Of Large Online Social Networks (OSN)

Social network data has two characteristics:the first one is the great scale of data. The number of users in the domestic and foreign popular social network platform is over one hundred million and the links between these users are more. Overall analysis of the whole network is unrealistic; the second one is the complex network structure. The relationship that contains a multi-level entity relationship between the whole network users is organized by all the users themselves. Based on the sampling method, it is difficult to restore and deal with such a complex internal coherence. Social networking internal multi-level entity relationship is the key problem that affects user sampling, measurement and evaluation in social network. This article give a exploratory research on internal multi-level entity relationship of social network hoping it can offer a better solution for user sampling, measurement and evaluation in social network.This paper mainly focuses on the research on asymmetric relations of large online social network. Social Network asymmetry mainly reflects in the node asymmetry, namely the imbalance of user influence and link asymmetry. This article will name the nodes that have advantage in the social network as the core node, namely the "Star Users" mentioned above, nodes that have disadvantage as the peripheral node. The author noticed that from the perspective of node gradation, the Social Network can be divided into three parts:Core Network, Peripheral Network, and Core-Periphery Structure. The Core Network is the key point of this paper.The third chapter of this paper discusses the Social Network node gradation and then selects Sina Weibo which is the biggest Social Network and have the most widely influence in China as the research object, building a35million Sina Weibo user’s network after data cleaning. First of all, through the statistical analysis, the author shows the network degree distribution and following ratio features, turning out that the degree distribution of Sina Weibo accords with typical power distribution; Secondly, the author find the Core Network that consists of core users who have more than5000fans from35million users in the network and use Degree Distribution, Following Ratio, Clustering Coefficient, Network Density, Edge Symmetry to analyze the properties of the Core Network in Sina Weibo; Thirdly, in order to validate the effectiveness of different sampling methods in detecting the Core Network, this paper analysis the original core network and two core network after Snowball Sampling and Random Walks, finding that Snowball Sampling has more advantages in the study of the Core Network; Finally on the basis of the previous work, design three experiments to analysis the effectiveness of Snowball Sampling in detecting the Core Network with different sampling seed, sampling depth and sampling ratio. As far as the author knows, this article is the first to focus on detecting the core of social network and take it as a character. At the same time, this article analyzes the coverage of core node, degree distribution and the density of fans network of core users, they reflect a lot of core network features.