A Two-Stage Machine Learning Approach to Forecast the Lifetime of Movies in a Multiplex

Abstract

Collecting over $2.1 billion annually, the cinema exhibition industry contributes 55% of the total revenue towards the Indian film industry. Selection of films is one of the most economically crucial decisions in cinema exhibition. Film selection is incredibly complicated to execute in India owing to its diverse demographic across regions and the resulting behavioral complexity. Working with data from one of India’s leading multiplexes, the authors offer a two-stage solution using machine learning to predict if a movie would proceed to be screened in the following week and the number of weeks it would continue to be screened if it does. The estimation of a movie’s lifetime helps exhibitors to make intelligent negotiations with distributors regarding screening and scheduling. The authors introduce a new metric MLE to evaluate the error in predicting the remaining lifetime of a film. The approach proposed in this paper surpasses the existing system of lifetime prediction and consequent selection of movies, which is currently performed based on intuition and heuristics.