Predicting Movie Success and Academy Awards through Sentiment and Social Network Analysis

Abstract

This paper introduces a new Web mining approach that combines social network analysis and automatic sentiment analysis. We show how weighting the forum posts of the contributors according to their network position allow us to predict trends and real world events in the movie business. To test our approach we conducted two experiments analyzing online forum discussions on the Internet movie database (IMDb) by examining the correlation of the social network structure with external metrics such as box office revenue and Oscar Awards. We find that discussion patterns on IMDb predict Academy Awards nominations and box office success. Two months before the Oscars were given we were able to correctly predict nine Oscar nominations. We also found that forum contributions correlated with box office success of 20 top grossing movies of 2006.