AUDIENCE BEHAVIOUR MINING BY INTEGRATING TV RATING WITH MULTIMEDIA CONTEN

Abstract

The TV broadcasting industry relies heavily on
TV ratings as a measure. Although television
ratings are mostly used for commercial
purposes, they may also serve as a sociological
indicator that represents people's interests.
Using data mining techniques applied to
television ratings, this article lays forth a
framework for understanding viewer habits.
With the framework we've built, we can now
use TV ratings to uncover a plethora of
audience behavior patterns. Various forms of
information, including the most popular news
programs and the most effective graphic
elements for achieving high TV ratings, may be
semi-automatically discovered when used in
conjunction with other multimedia items like
text and video.
By zeroing in on the times when the rating data
shows a significant shift, or when a large
number of viewersturn the TV on or off, we can
learn how audiences behave. From multimedia
materials, we glean detailed descriptions of key
points; next, we use a number of filtering
algorithms to isolate relevant patterns. Several
examples of using this framework to find new
information shown that it can successfully
extract different kinds of audience behavior. As
far as we are aware, this is the first study to
examine ratings data with video and other
multimedia data.