Dynamic Bayesian predictive model for box office forecasting

Abstract

Film industry plays a vital role in driving economic growth in the modern society. Though financial gain of a successful movie can be fabulously huge, competition in the realm is intrinsically competitive. Box office forecating of a movie therefore becomes significantly pivotal as movie producers can properly allocate funds and resources in producing and distributing a movie. Conventional methods statistically model the correlation of box office and indicators such as classic movie attributes or post-release information which often lack real-time efficiency. The proposed method in this paper uses a dynamic linear model with a Bayesian framework, for an improved performance in predicting daily box office. The method considers both prior knowledge from big historical data and the daily refreshed data for dynamically updating model coefficients. Experimental results prove efficacy of the proposed method on a sample data of movie box office revenue.