Research On User Feature Analysis Technology Based On Social Network Hot Events

With the development of the Internet,more and more users express and spread their opinions on hot events through social network platforms such as microblog.Social network hot events refer to events that attract public attention and have strong response through social network.In the spreading process of hot events,it will arouse public opinion,produce large-scale chain reactions,and even endanger society security.As the main participants in hot events,users are the core of social media,and play an important role in the process of generation,evolution and dissemination.Therefore,how to effectively analyze the user feature of social network in hot events becomes the focus of current academic research.User feature analysis is a hot topic in social network and data mining,which involves a wide range of research.Based on the user sentiment and user influence,the key technology of user feature analysis is studied in this paper.The analysis of the user sentiment is a study of user's views and attitudes,which can be used to help government department understand the public opinion impact of hot events in time.The analysis of the user influence is a research on the importance of users and the ability of information dissemination,which is of great significance to the supervision and guidance of network public opinion.The main work is as follows:(1)Aiming at the problem of sparse feature and poor topic focus of the text,a sentiment classification method based on local word vector feature and global topic feature is proposed in this paper.User sentiment is measured by the global document topic feature and the local word vector feature.First,local word vector feature is trained based on neural network by using statistical features,sentiment dictionary and distributed word vector.Then,global topic feature is obtained by introducing topic model into feature extraction and fusing neural topic model.Experiments show that the combination of global topic feature and local word vector feature can enrich the feature of neural network and effectively improve the accuracy of sentiment classification.(2)Aiming at the current user influence research focusing on user factors such as social structure and user behavior while ignoring the quality and content of microblog,a user influence method based on user information and content information is proposed in this paper.User influence is measured by the features of direct influence,indirect influence and microblog content.Firstly,based on microblog users' information,a multi-angle model based on prestige,activity,spread and innovation is established to describe direct influence.At the same time,the idea of PageRank is used to get the influence of fans.Then,based on content and comments,the content influence is obtained by integrating users' sentiment tendency and analyzing the text semantics.Experiments show that the proposed model can measure the user influence in hot events more effectively.In summary,based on hot events in microblog,user sentiment and user influence are studied,and the effectiveness and accuracy of the proposed method is proved through experiments,which plays an important role in improving the efficiency and accuracy of public opinion monitoring and guidance.