Research On Key Technologies Of User Feature Analysis For Social Media

With the continually maturation and development of Web2.0, social media has become an important part of people’s lives. Social media is an online interactive media that has an extensive user involvement. It allows users to publish and disseminate information online to communicate, cooperate with each other, and form virtual online communities. Blogs, micro-blogging, wiki, forums and podcasts are all belong to social media. The focus of social media is "user". So how to analyze the characteristic of network users accurately and effectively is becoming a hot spot of academic research and application development. In social networks, there is a great practical value in carrying out a proper evaluation for users and then executing the appropriate measures for them. For example, finding out the "opinion leaders" regarded by netizen and then monitoring the changes of public opinion in order to guide the orientation of opinion; to infer the user’s personality and hobbies is useful to promote products and inculcate the concept, and so on.From the angle of complex network analysis and data mining, this paper mainly focused on key technologies of user feature analysis for social media. Author has put emphasis on key issues including "user influence analysis","user role recognized","evolution of user’s community" and so on. Among them, the user influence analysis is the study of the important degree and the information dissemination ability of users, whose results could be used to define the user role and monitor the changes of public opinion. User role recognition divides users who have the similar behavior characteristic or social position into the same category, then studies the difference and effects between roles and individuals in network. The community is composed of individuals that have close relations. The community evolution, which takes user as the node, makes research on the changing reason, the tendency and the influence of the closeness of users’ relations. This article launches the research according to users’ roles and the changing situations of network architecture.The main content and innovation of this paper are as follows.1. Proposes a user influence ranking method based on the tendentious changes. The existing influence sorting methods exist the following problems:1) The effect for the real network whose relations are sparse is poor;2) Doesn’t support dynamic updating;3) Ignores the hidden tends or emotions in the generated content of users;4) Never consider that users may accept or resist the influence of others selectively. Therefore, this article has studied the ability that users influence others and the extent that affected by others from the angle of tendentious changes during the process of user interaction. Also, this paper has proposed a user influence sorting method based on the tendentious changes. Firstly, this method metrics the influence of posts using static property value, spread value and tendencies value. Secondly, decomposes the complex user interactive process into multiple local reply chains and increase the indirect reply relationship. Thirdly, judges that whether one user has been influenced by another one and the influence degree according to the tendentious changes situation of analytic posts in each partial reply chain. Finally, people could get users’ influences and ranking at all levels quickly by query and supports the dynamic updating of data. The experiment shows that the structure method of local reply chain is rational and proves that the influence sorting method proposed in this paper is not only effective, but also could find the high influence users in some other light.2. Proposes a user role recognized method based on directed topological potential. Topology potential theory has been applied to measure the importance of undirected network node, while most of the various systems in the real world are abstracted into directed weighted networks. Also, the calculation process of node topology potential involves the shortest path problem, so it has high computational-complexity. This article gives the definition of in-degree topology potential and out-degree topology potential based on directed weighted network and proposes one kind of topology potential algorithm based on the influence scope to reduce the computational complexity. According to topology potential distribution of all nodes, this article proposes a new user role analysis method, which provides a favorable support in studying the features of roles and users, detecting dynamic network evolution events and the like. The experimental results show that the topology potential has good effect on measuring the importance of node and dividing the roles; the sudden events of dynamic network can be identified based on the role changing and the differences of important ranking.3. Proposes a social network evolutionary method based on the role of nodes. In the existing study of user roles, many scholars have defined the number and characteristics of roles that have achieved good results on a particular dataset. But there are two problems:1) The generality is poor and must reanalyze if the dataset has been replaced;2) In the real world, user’s behavior and relationships are complicate and user roles are varied. It’s very difficult to describe and identify them by the artificial definition. So, this article proposes a user role found algorithm based on the tensor decomposition model. This algorithm not only can set the number of roles automatically, but also can reflect the behavior characteristics of role in specified period of time. Furthermore, this article extends user role to community roles and raises a community evolution analysis method based on the distance of community roles and the node overlapping. The experimental results indicate that the behavior characteristics of the identified roles are consistent with the fact and the community evolution analysis method proposed in this article has better effect than comparison algorithms.Overall, this paper is mainly faced to the social media and has conducted a thorough analysis and research on user features based on the aspects of user behavior and relationship.