Research On The Emotional Tendency Of User Groups Based On Social Network Analysis

With the rapid development of Web 2.0 era,cyberspace has increasingly become an open and free public opinion field for users.Compared with the real world,the cyberworld also has the characteristics of secrecy,diversity,non-authoritativeness and flatness.Network users can freely express their personal opinions and opinions on social life in various social media,such as micro-blog,circle of friends,forums and posts.Bars,e-commerce evaluation websites and so on have become a platform for users to disseminate massive information,which contains rich public comments and emotional information.To a certain extent^they can form social public opinion and affect the development and trend of emergencies or social public events.Therefore,the judgment of users'emotions can not only reflect the public's emotional trend,facilitate the monitoring of network public opinion situation,and achieve social stability,but also play a huge role in various micro-fields,such as online product review analysis and evaluation,stock market sentiment analysis and risk control,film style analysis and prediction of box office revenue.However,in the process of Internet content dissemination,the phenomenon of Balkanization appears in the real world,that is,the individual network users gradually evolve into groups with the same interests or emotional tendencies.Traditional affective analysis methods have achieved some results in text feature extraction,but neglected the user's own social relations,assuming that the data samples are in line with the assumption of independent and identical distribution,and only analyzed the emotion of a single user.There must be universal association among users.The overall emotional orientation of users'social groups is more obvious and stable than that of single users.Therefore,it is of great significance to study the emotions of users'groups from the perspective of social network analysis.The research work carried out in this thesis includes the following aspects:First,social network association mining.In this thesis,a FRAP-based network community partition method is proposed from the point of view that similar user nodes have similar group tendencies.Firstly,the location information of user nodes is depicted at the level of vector space,and the energy function is defined into the force guidance space.The convergent position vector is obtained by decreasing the energy size continuously,and the multi-level neighbor information is fused into the node difference matrix.Then,the core element nodes are clustered by the confidence value transfer of the difference degree,and the core element nodes are obtained by the label optimization after clustering.The local convergence of the current module values is achieved.Finally,the class labels of the nodes are aggregated and updated by the labels of the measurement coefficients,and the result of community division is obtained by the historical label sequence generated under the guidance of the attribution values.Experiments show that the FRAP method can achieve better performance in the test set.Secondly,the analysis of user-oriented emotional orientation.From the perspective of user groups,multi-granularity evaluation of emotion is carried out.Firstly,the research hypothesis based on single user's emotional preference is put forward,and the sub-hypothesis about user's emotion is validated by establishing the first-level indicators of emotional intensity and diffusion heat.Then,the law of positive emotional bias such as joy and emotional diffusion in comment data is analyzed;secondly,the emotional inclination oriented to user groups is given.To the analysis method,the closely related user groups are established by the method of community partition,and the change of user groups'emotional trend is analyzed from the perspective of diachronic and synchronic.Based on the scale,density,relevance and other related attributes of user groups,the discussion on the selection of hyper-parameters and the mining of user group's emotional tendency characteristics are given.