Internet User Portrait And Fake Information Related Feature Mining

Due to the fast spread,low cost,and easy access of Internet information,it provides users with a convenient channel for obtaining information and also provides a convenient and fast channel for the widespread dissemination of fake information.Internet fake information often misleads users,distort truth,Guiding and spreading bad emotions even stimulates a strong public opinion reaction,which has extremely adverse effects on individuals and society.Compared with the fake information in traditional media,fake information in the Internet social network has formed a unique display form and spread effect,and has more group,concealment,and accuracy of dissemination.Its destructive power and influence are much higher than before.The detection and prevention of fake information on the Internet has important research and practical significance.Traditional researches on the detection of fake information on the Internet are mostly identification and detection of the authenticity of the information content,which cannot effectively monitor and analyze the fake information from the key nodes of the generation and dissemination of Internet users.The purpose of this study is to use user portraits as a means of describing Internet users,to extract features of Weibo user portraits based on Internet information big data,and to use user explicit features and speech data to mine implicit features that are not public or recorded by users.Feature information complements user portraits.Based on the Internet user portraits,user characteristics and detection methods related to fake information on the Internet are explored.The main work of this article is as follows:(1)This paper proposes to complete the basic Weibo user profile construction and explicit feature extraction of Weibo users based on the basic information dataset of Weibo users.Based on the explicit features,K-Means clustering algorithm is used to divide Weibo users and mine them.The implicit characteristics of the user category finally divided the users into advertising,ordinary,star and active users,verifying the feasibility of mining the user's undisclosed implicit characteristics based on the user's display characteristics.(2)This paper proposes a method to mine the implicit characteristics of Weibo users' emotional tendencies and personality characteristics using Weibo user speechinformation.In this paper,Fast Text and BERT sentiment classification models were trained based on sentiment classification data of Weibo,and the accuracy and efficiency of the model were compared.The hidden features of Weibo users' emotional tendencies were mined and obtained.The Pear model was used to mine the personality characteristics of users and obtain Weibo The user's big five personality score is used as the implicit personality of the user.In the end,two user implicit features were used to improve Weibo user portraits.User implicit feature mining based on Weibo user speech was realized,and the effectiveness of the proposed method was verified.(3)This article completes a set of micro-blog fake information user opinion data labeling,and proposes that based on the labeled data,the correlation between the perfect micro-blog user portrait and the fake information opinion spread by the user can be mined through correlation analysis and statistical analysis.It is concluded that negative emotional users and open personality users are more likely to believe fake information.(4)Based on the experimental data and procedures in the research process,this paper uses a three-tier architecture to design and package a set of Internet user feature mining analysis framework,and the output results are displayed in the form of data tables.The data mining analysis framework is provided in the form of a Python library,which can be transplanted to different systems and platforms,supports continuous research on user characteristics,and can effectively improve research efficiency.