Design And Implementation Of Digital TV Audience Rating Analysis And Forecast System

Nowadays, television, a powerful mass broadcast media, is one of the threemainstream advertisement providers. Along with the development of contemporarystatistics, TV ratings survey has become very important for both media andadvertisement. The statistical results from rating surveys play a key role in TVprogram producing, in selection of broadcast contents and AD providers. Graduallythe data of TV ratings survey has turned into an authoritative standard and a key indexfor the survival of a TV program. It can not only evaluate the quality of programs,guide their development, but also evaluate the whole channel and its advertisementvalue. However the current status of rating survey is far from satisfaction. Becausethe surveys carried out by most TV companies only sampled simple data, which areinsufficient to cover all aspects of the program so the evaluation is not quite accurate.Furthermore, present rating system can only provide a posteriori analysis, predictionis not possible yet.To address these issues, we developed an audience rating analysis and forecastsystem to investigate various rating factors. The main theme of this thesis consists ofthe following parts:(1) Based on the data gathered by the two most authoritative surveyorganizations, CVSC-Sofres and AC Nielsen, a business process and pattern wasestablished for the TV ratings analysis and forecast system.(2) For an accumulated database on the scale of Gigabytes or even Terabytes,regression method is applied to reduce its size. The model is built by data miningtechniques. And a data warehouse is established for the huge amount of CSM data.(3) The system is built on the J2EE platform. The framework is constructed bymeans of the Struts MVC and AJAX techniques. The commonly used methods in thedata mining community like classification and prediction are adopted to analyzevarious rating factors and reveal the relationships among them.Our experience shows that the current rating analysis and forecast system hasturned out to be a fine achievement. By analyzing such factors as time slot, brand,channel, etc., an accurate evaluation system is established, which enable us to quantitatively predict a proper program arrangement and a wise advertisementinvestment.