Early-stage Prediction of International Revenue of Bollywood Movies—Emphasis on Simultaneous Over Sequential Release

Abstract

Bollywood movies made in the Hindi language are regarded as one of the most popular Indian movies. Bollywood movies like other Indian movies are successful in retaining their characteristics, which is significantly different than Hollywood movies. Overseas earnings contribute a significant pie of Bollywood movies’ total revenue. International revenue prediction requires a new approach due to the changing release strategy by the investors. Movie investors worldwide are moving away from sequential release first in the domestic market followed by release in the international markets to simultaneous release both in the domestic and the international markets. Revenue prediction at an early stage before committing money to the project is more valuable than prediction at a later stage just before or after the movie’s release. We utilize multiple machine learning algorithms to improve the baseline prediction accuracy significantly. Extreme gradient boosting, the best algorithm, reduces the baseline prediction error by 18.05%. Adopting different algorithms to different scenarios improves prediction accuracy relative to applying one algorithm across all scenarios. As an example, robust regression generates the highest prediction accuracy for movies with higher star power, whereas extreme gradient boosting achieves the highest prediction accuracy for movies with lower star power. More accurate prediction before committing money to the movie project strengthens investors’ judgement and reduces investment risk. Our study addresses a new trend of simultaneous movie releases in domestic and international markets and predicts international revenue without knowledge of domestic performance.