Micro-blog Based Users Latent Attribute Recognition

Micro-blog's user attribute recognition is one of the most important research fields in data mining.With the rapid development of the Internet,people share the convenience of the Internet at the same time began to pay more attention to micro-blog.Micro-blog is a social network based on the platform,because of its timeliness and randomness characteristics,can always reflect the changes of things around,as a human interaction and communication essential tools.Due to the excessive number of users,micro-blog will produce a large number of data,how to dig out the effective information from these data has become an important issue in today's society.At the same time,it is still a key link in the research of social media that how to analyze the data of micro-blog users,deduce the behavior characteristics of users and detect the network security issues.For micro-blog users of gender,age,region,interest preferences and other characteristics,how to effectively predict,will be able to avoid some of the mistakes,and even benefit the whole society.In today's highly developed science and technology,for the processing of micro-blog data,can not be separated from the application of computer technology.The micro-blog data in the online media to show the characteristics of a variety of gestures.How to understand the basic information of micro-blog users,from which to identify the user attributes of micro-blog is its core technology.Although studied at home and abroad has been on the micro-blog data mining,but the main research contents of micro-blog emotion analysis,social groups based on mining,topic detection and research direction,the user attribute relative less,such as gender,age,user area classification etc.In this paper,through the two aspects of gender and age of micro-blog users,to introduce the micro-blog user attribute recognition.(1)Age estimation of micro-blog users based on support vector machines.By training SVM(Support,Vector,Machine)classifier,the method of improving feature weight is improved,and different feature vectors are constructed to improve the accuracy of age estimation for micro-blog users.(2)Gender classification of micro-blog users based on topic model LDA.On the one hand,through the modeling of the micro-blog user interest,released by micro-blog combination of user content and attention behavior,establish content preferences and preferences for attention,structural characteristics of gender classification,theexperimental results show that the effectiveness of micro-blog users interest preferences than traditional lexical category features;on the other hand,according to micro-blog influence activity different users which brings to the classification results.The experimental results show that using micro-blog users' attention preference feature(not relying on micro-blog's text content)to infer classification is more effective and robust.