TV audience rating is an important indicator regarding the popularity of programs and it is also a factor to influence the revenue of broadcast stations via advertisements. Presently, the only way for assessing audience rating is the Nielsen TV rating, which depends on a small number of randomly selected representative groups, because of practical considerations such as cost and survey time. The way to obtain audience rating is using 'People-meter' which is a device installed in user's house and regularly records the rating surveys. However, we are not able to know the audience rating immediately since sometimes we have to make a marketing decision and lack of indicator. Currently, the present media environments are drastically changing our media consumption patterns. We can watch TV programs on Youtube regardless location and timing. And Nielsen TV audience rating does not take the social networking site into account. In this paper, we develop a model for predicting TV audience rating. We accumulate the broadcasted TV programs' word-of-mouse on Facebook and apply the Back-propagation Network to predict the latest program audience rating. We also present the audience rating trend analysis on demo system which is used to describe the relation between predictive audience rating and Nielsen TV rating.