A Deep Learning Approach to Predict Film Box Office in the Chinese Domestic Market Based on Feed-Forward Neural Network

Abstract

With the booming development of the Chinese film market, China has become the world's second-largest film market and the main engine for industry development. In order to improve the efficiency and predictive power of movie box office prediction in the Chinese domestic market, this study proposes a Feed-Forward Neural Networks model with two and three hidden layers by predicting the box office of 478 films in the Chinese market from 2019 to 2022 using movie metadata, post-release rating data, social media related big data and theaters arrangement. Loss curves for the training and validation sets are presented for model comparison with 400 films being used as the training set and 78 films being used as the validation set. The results show that the Feed-Forward Neural Network with three hidden layers has greater fitting and predictive power in model generation, enabling it to effectively predict the movie box office.