Predicting Box Office and Audience Rating of Chinese Films using Machine Learning

Abstract

The quality of a film can be judged by various measurements, with box office (BO) and audience rating (AR) being two of the most important, that are influenced by different factors. Examining the influence of these factors on the film's quality can provide clear indications regarding the factors that affect the strategic decisions of film production and distribution. The differences in cultural backgrounds and the film types, however, makes it difficult to determine the relation between these factors and BO/AR as the majority of these factors are difficult to assessor measure. Consequently, it is almost impossible to predict BO/AR before the film is released. This paper studied the relationships between BO/AR and film elements that can be easily obtained before the film's release. Both type-independent and dependent correlation analysis have been conducted and machine learning models were designed to estimate BO/AR. To conduct such studies, a database consisting of 34 Chinese films has been established. The analysis and experimental results demonstrated that despite the correlation between film elements and BO/AR differs according to the various types of films buy the proposed model can accurately predict them using film elements data.