User Hidden Attributes Inference And Attributes Cluster Analysis Based On Social Media

With the rapid development of the Internet,social media spread rapidly,and its influence is not as same as before.With the help of AI,we can deeply understand the basic information of social media users.By mining the potential behavior patterns and basic characteristics of social media users,providing various and personalized decision support for government departments,and solving practical problems have become a hot topic of common concern in academia and industry.However,the complexity of social media is not taken into account in the current research.Only a single blog information and traditional model can not fully describe the hidden attributes of users,which leads to the unsatisfactory effect of the hidden attribute analysis and user interest mining model.For this reason,this thesis introduces a social media user hidden attribute analysis model based on multi-features and a social media user interest mining method to solve the above problems.And three kinds of hidden attributes of users,age,gender and interest,are studied based on Weibo platform.The main work of this thesis includes:Firstly,the existing methods of user hidden attribute analysis use single user information and traditional models,which leads to the problem of low accuracy of user hidden attribute analysis.This thesis focuses on two types of user hidden attributes: user age and user gender.This thesis improves the stacking model on the basis of building a multi-feature system of microblog users.First,the text features are constructed from the Weibo text data by word2 vec model;then the basic features for Weibo hidden attribute analysis are constructed from the Weibo user data,and the composite features are constructed according to the basic features;finally,the improved three-layer stacking algorithm is used to construct the Weibo user hidden attribute analysis model.The experimental results on Sina Weibo dataset show that this method can effectively improve the effect of hidden attribute analysis of Weibo users.Secondly,in order to mine the interests of social media users more accurately,this thesis takes Weibo platform as an example to cluster the LDA topic model of Weibo users' texts,so as to obtain the topic characteristics suitable for user interest mining.On this basis,using xgboost algorithm to build user interest mining model with the combination of the user feature of Weibo information and the personal feature.Experiments on Sina Weibo dataset show that this method can improve the effect of user interest mining.Thirdly,this thesis designs and implements a Web system to show the model by using Django framework and Java Script,and adds crawler to realize the real-time acquisition of Weibo user data.Therefore,the Web shows the analysis results of Weibo user hidden attributes.