Research On The Statistic And Analytic Methods Of Digital TV Viewing Data

With the rapid development of digital television technology,the media market is booming,the digital television ratings index as one of the important standards of market analysis and program evaluation,which plays a vital role.Through the analysis of the figures,we can observe the viewing behavior of people,then promote programs and advertise.How to predict the audience rating of a program accurately,which benefits the enterprise users,advertisers and television station.This paper has analyzed the viewing data,combined the existing ratings statistical indicators,designed a functional digital television viewing statistics system,and uses the data mining algorithms to predict a program’s ratings.The main content is as follows:(1)Combined with the statistical indicators that are relative to digital television viewing figures,the structure and principle of the digital television viewing statistics are analyzed in detail.The algorithms of data mining and relevant document literature are researched deeply.At last the design scheme of digital television viewing statistics system is given.(2)Design and implementation of the digital television viewing statistics system.The system is divided into two parts:the front desk uses B/S network architecture,which is realized on the.NET platform,the function contains DVB channel time analysis,DVB channel programs analysis,VOD,rating prediction,knowledge base etc.The background combined with the corresponding statistical indicators analyze the viewing data and transmit the result to the front desk.The system uses the SQL database(SEVER2005)to design the relevant schedules,real-time viewing data table,VOD ratings data table,the prediction information table etc.(3)Analysis and implementation of the prediction algorithm relative to viewing data.Through the research of data mining prediction algorithm and the influence factors of program’s ratings.The paper selected the art theme,famous brand etc 15 factors which influence the program’s ratings and then fuzzy quantitative evaluation of these factors,and adopted the decision tree algorithm and the BP neural network to predict the program’s ratings.(4)The prediction algorithm verification and the test of the digital television viewing statistics system.Due to the confidentiality and privacy of the viewing data,there is no open source viewing data is available for the system test,only by searching the websites for audience rating,and simulate a user’s daily viewing habits as real as possible,then verify the feasibility and practicability of the system analysis based on these data,the test result is satisfactory.