Prediction of box office revenue of movies

• Main Author: Lim, Joey Jiayi
• Other Authors: Li Boyang
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Box office prediction
• Online Access: https://hdl.handle.net/10356/175059
• Institution: Nanyang Technological University

The volatile nature of the film industry has rendered box office prediction a challenging task. Factors like social media buzz and ever changing trends have left film producers and studios scrambling for a foolproof method for predicting box office revenue to mitigate risks. Leveraging a rich and comprehensive dataset comprising information relating to 36,000 movies sourced from popular movie sites such as TMDb and IMDb, this study aims to analyse the importance of feature engineering on the accuracy of predictor models. Apart from features such as budget, this study explores the use of sentiment analysis to determine audience reactions to the trailer of a movie. Four distinct models: Random Forest, LightGBM, Feedforward Neural Network (FNN) and Recurrent Neural Network (RNN) were analysed and their performance compared against each other to determine which model best suits the task of box office prediction.