Research On TV User Profiles Based On Publish Micro-blog Data And Viewing Behavior Data

With the rapid development of TV technology and Internet,the traditional TV.that only receives signal has changed dramatically.Nowadays,people can watch many kinds of TV programs at home by using smart TV or by installing bidirectional set-top box.And at the same time the TV service provider can also get all the accurate operating data of TV users in time.In order to provide TV users with TV programs that better satisfy their appetites and advertisements that meet their needs,it is of vital importance to construct precise user profiles by analyzing data of their TV-watching behaviors.The traditional method of constructing TV user profiles is obtaining the label of TV program from the Electronic Program Guide(EPG),and then obtaining the TV user profile labels by analyzing the relationship between TV user and TV program.However,the traditional method has some limitations because of the strong subjectivity of the EPG label and the narrow coverage of the label.In this paper,we studied how to use public micro-blog data to improve the accuracy of TV user profiles we construct.The core part of the research are we finding the relevance between micro-blog users and TV users,and then using the accurate micro-blog data to predict the TV user profiles.Specific tasks of the paper include the following parts:1)This paper introduces the background and significance of TV user profiles,researches and confirms that some micro-blog users have interests in TV programs and we can also get accurate micro-blog user profiles information;2)Then,a highly effective web crawler is designed to obtain data of those micro-blog,and then Chinese word segmentation algorithm,configuration dictionary and TF-IDF method are used to mining features of those TV programs that micro-blog users concerning about;3)Then a large amount of TV users' operating data are cleaned,and we manage to get features of those TV programs that audiences pay close attentions to;4)We associate the micro-blog user with the TV user and design user profiles tag set,and for each tag we use the micro-blog data to construct a classification prediction model,then those models are used to predict the TV user profiles;5)In order to evaluate the accuracy of user profiles,we use a month of real TV users' operating data in a city,and our method evaluated by the content-based recommendation system,and the effectiveness of the method of obtaining TV user profiles based on micro-blog data is tested.Experimental results show that our method has better performances in Precision and Area Under Curve(AUC)compared with other algorithms.