Extractioin And Analysis Of Relationship Between Network User Virtual Identity

With the popularity and rapid development of the Internet, the way that people communicate with each other changes from paper letter to e-mail, Wechat, QQ, and Weibo. Accordingly, a new concept--network user virtual identities (NUVIs) comes into being. In a broad sense,network user virtual identity is an identity that can identify an individual.In a narrow sense, it is an account that user applies for in different applications. Compared with real life, these NUVIs, which maybe incomplete and unreal, represent an identity of an individual in network.People communicate with each other by two NUVIs more and more closely and frequently on the mobile Internet, that will form a relationship pair. This relationship pair maybe directed or undirected. Large-scale relationships pairs will form a large-scale social network, due to their intensive relations with each other, such as social network in real life. So,this thesis will firstly design a complete extraction system to extract NUVIs from raw network packet flow; then analyze the relationship strength of these NUVIs. Considering the scale of these relationship pairs,we will use Spark and GraphX to implement our relationship mining algorithm.This thesis firstly introduces basic concept of network user virtual identity and analysis of social network, and relevant theories of complex network, community detection, label propagation, and GraphX. Then we design and implement a complete NUVIs extraction system that can acquire relationship pairs from raw network data packet flow. We will propose a method to analyze these pairs. From bigger angle, we parallelize the traditional COPRA algorithm to detect communities in these relationship pairs. From smaller angle, we implement Two-Hop and common friend algorithms to acquire relationship strength between these pairs. And we will get any relationship strength of two NUVIs. Finally,we summarize the research contents and prospect for future researches.