| With the fast development of computer and internet technologies,various online social networks constantly emerged.Online social networks have the characteristics of freedom to access,easy to use,fast information propagation speed and high interactive and have attracted lots of attention.Nowadays,online social networks have become one of the most popular information platforms in people's daily life and work.Based on the user interactive behavior like retweet and comment,the virtual relationships between users can be regarded as the expansion of social ties in real world to a certain extent.To study the phenomenon and law in online social networks are very meaningful to solve the related problems in the real world.As the core of online social networks,the characteristics of users are close related with topology structure of online social networks,user type classification,information diffusion principle and some other important research.Therefore,how to effectively utilize user characteristics to carry on the application in fields such as public opinion monitoring and precision marketing has great value.We have used the real online social network data as research object and carried out detailed research on some important research problems by user characteristic studying and analyzing.Our study includes user influence evaluation,user type classification and population health research with online social networks.Among these research fields,user influence evaluation and user type classification are always the focus of online social network studies.The related studying achievements can make public opinion monitoring and marketing more efficient.Population health research with online social network is one of the hottest research issues in recently.Online platforms can provide new ways for the population health problems with data and methods and can do some useful supplement to existing methods.The main content and research results are as followed:Firstly,based on user characteristics like topology structure,user behavior and user information we have proposed a user influence evaluation method with multiple characteristics.This method aims at how to effectively use the characteristics to evaluate user influence.We use Bayesian network to synthetically measure the effect of various characteristics on user influence.And then,learning from PageRank algorithm,the adjacent factor on user influence is also considered.Therefore,this method can evaluate user influence in a comprehensive way and can void the disadvantages of existing methods such as insufficient influence consideration by the single index methods and physical interpretation confusion by the simple weighted methods.Finally,the experiments are carried out with Sina Weibo data and the results demonstrate the validity of our method.Secondly,based on the analysis on information retweet chains among online social network users,we have proposed a regional interaction model.This model can describe the interaction between different user with different distance based on retweet behavior which can truly reflect the real interactive mode of online social network users.Then based on regional interaction model we have studied user type classification in online social networks which can make the result more reasonable according to the user behavior and influence mode.With the regional interaction model,user influence can be divided into direct influence and indirect influence and users can be classified as different categories such as influential users,normal users and unusual users based on these two kinds of influence.The experimental results show that the regional interaction model can effectively identify multiple user classes in online social networks.Compared with the existing methods,our method is more efficient.Thirdly,we have carried out exploratory research on population health status with the information characteristic of online social network user and proposed an obesity status of population predication method with analysis results.In online social networks,users may post large amounts of information containing user life style,hobbies,interests,emotion and physical status.The health status of people in real world can be inferred with these kinds of information by mapping from the virtual world.In online social networks,the scale of information and its openness provide a substantial data platform and new way to solve the existed health research problem.Compared with the methods using traditional data from medical institution,methods with online social network data are more efficient.In this paper,we have analyzed some kinds of user characteristics related to obesity based on the existing research and extracted relevant information from online social network users.Then we have analyzed the relationship between user characteristics and obesity rate in different areas.we have utilized these characteristics obesity-related as the coefficient of obesity group variation and predicting the obesity trend in different areas.The experimental results show that the proposed characteristics are close related to population obesity.Therefore,the obesity trend predication method based on these characteristics has certain validity and application value.At last,we developed a system on the basis of our research achievements which can be used to analyze online social network users.This system can effectively identify influential users,abnormal users in Sina Weibo and can reveal the obesity status in a certain area.The function of this system is close related with the user influence evaluation method and user classification method proposed in this paper and has some practical value.This paper is funded by National Natural Science Foundation of China(NSFC)"Research on information propagation dynamic analysis and modeling in heterogeneous online social networks,(item number:61172124)" and "Research on regional population health factor analysis and evolution laws based on diversity and heterogeneity big data,(item number:61571360)". |
PDF Full Text Request