Predicting Movie Ratings from Audience Behaviors on Movie Trailers

Abstract

Movie rating is a measure of viewer’s reaction to movie performance at the box office and also a key feature to garner publicity. Movie ratings are feedback measures given by a subset of the audience voluntarily. If the degree of effect on the human mindset can be measured through real-time behavior analyzing and rated, the results can help film houses to understand the secret of generating a commercial success movie. Prediction of movie ratings is a complex problem. Viewers, producers, directors, and production houses are curious about how a given movie will perform in theatres with different customer segments. Research works have been carried out relating to movie rating prediction using social networking, blogs articles, but much less has been explored by the consumer behavioral data and attributes while watching a movie continuously and using emotions and body movement dimensions [1 – 4], [7], [12]. We created an audience footage data set and transformed it into numerical feature data representing the audience’s behavior. Prepossessing and machine learning approaches were applied to build an efficient model that can predict the movies’ popularity.