Research On User Influence Model Construction And Personality Traits Performance Under Strong Connection

With the development of mobile technology and online social platforms,online socialization has become an indispensable part of people’s daily life and work.Moments has become a platform where users share information,show themselves,strengthen contact and interact.According to the strength of social relationships,social software can be divided into weakly connected social software and strongly connected social software.Users are not closely related to each other and do not have strong emotional connections on weakly connected social software,such as Weibo,TikTok and various forums.The relationship between users on strong connection social software migrates from offline social relationships,so users are closely related to each other,such as WeChat,QQ and etc.With the rise of social e-commerce,the focus of social network research has also shifted from opinion leaders in weakly connected networks to opinion leaders in strongly connected networks.Opinion leaders in a strong connected network are excellent store owners or potential shopkeepers,they are important nodes in viral marketing.User influence measurement is the basis of opinion leader mining,but the research on the influence model of strongly connected networks is not sufficient,most influence models focus on the study of weakly connected networks.In our daily life,the influence of users in a strong connected network is mainly reflected in three dimensions: social links,interactive behaviors and post content.In addition,because the strong connected network is the offline social relationship migration,users are more familiar with each other.The evaluation of others,personality differences,social identity and personal likes and dislikes will be reflected in the strength of user influence.Therefore,this paper takes the WeChat moments dataset as the research object,elaborates how to construct the user influence model of WeChat moments,and analyzes the performance of personality traits in user influence.In summary,the research content of this paper can be divided into two sections.Section one explores the user influence model of WeChat moments.Firstly,this paper summarizes and analyses the research status of user influence research at home and abroad,then compares the advantages and disadvantages of three kinds of user influence measurement models(influence measurement models based on topology network structure,influence measurement models based on user interaction behavior and influence measurement models based on post content).Then this article crawls the WeChat cache database and extracts social link relationships,interaction datasets,and UGC datasets.Besides,aiming at the problem of divergent social networks caused by WeChat’s privacy protection,this paper proposes to adjust the network structure through link prediction,transform the divergent topological network into a reticular network structure,and adjust the influence measurement formula of behavior dimension and content dimension.After that,according to the information entropy of each dimension’s influence,the author uses the entropy weight method to calculate the weight distribution value,and obtain the initial user influence matrix.The comprehensive user influence is obtained through the iterative process of LeaderRank.Finally,the evaluation results are compared with the LeaderRank model,User popularity and User influence evaluation method based on topic and node attributes(UIEM)through repetition rate,opinion leader mining accuracy and recall rate.After the experimental analysis,the author finds the WeChat moments user influence model proposed in this paper has a good performance.This influence model enriches the user influence research area,provides reference for user influenc research on strongly connected social network,and can be applied to the social e-commerce shopkeeper mining,viral marketing and other fields.Section two discusses the performance of personality traits in user influence.Firstly,this paper reviews the current research status of personality traits from the perspectives of personality trait language-behavior feature analysis and personality trait assessment.Then,the language-behavior feature vectors are extracted from the dataset as independent variables,and the results of volunteers’ personality scale reports are used as dependent variables to model fit training.Using MAE,MSE and R2 as evaluation indicators,this paper compares the predicted results of three regression models: multiple linear regression(LR),decision tree regression(DTR)and random forest regression(RFR).The author selects the random forest regression to complete the remaining users’ personality trait assessment.Finally,the user clusters are divided into five trait clusters according to the users’ representative personality.On the basis of the section one,this paper analyzes the manifestation of personality traits in the user influence.Exploring the influence performance of personality traits not only proves the correlation between personality traits and user influence,but also enriches the internal perspective of user influence research,which provides suggestions for exploring the personality traits functionary mechanism and constructing personalized user influence model.