Predicting Box-Office Markets with Machine Learning Methods.

Area of Science:

• Economics
• Data Science
• Machine Learning

Background:

• Accurate prediction of gross box-office markets is crucial for investment and management in the film industry.
• Existing methods may not fully leverage diverse economic factors for precise revenue forecasting.

Purpose of the Study:

• To propose and evaluate a machine learning-based method for predicting country-specific movie box-office revenue.
• To compare the performance of eight different methods using various economic factors.

Main Methods:

• Employed machine learning, specifically comparing eight diverse methods.
• Utilized economic factors in combination with time-series forecasting experiments.
• Focused on support-vector-machine (SVM) models, particularly those incorporating Gross Domestic Product (GDP).

Main Results:

• Achieved a relative root mean squared error (RMSE) of 0.056 in the US and 0.183 in China (2013-2016).
• SVM using GDP showed the best prediction performance, utilizing readily available economic data.
• Validation in 2017 yielded error rates of 0.044 (US) and 0.066 (China); 2018-2019 predictions had mean relative absolute percentage errors of 0.041 (US) and 0.035 (China).

Conclusions:

• The proposed machine learning strategy, particularly SVM with GDP, is effective and versatile for movie box-office revenue prediction.
• The method demonstrates high accuracy and efficiency in both training and validation datasets.
• This approach offers a valuable tool for investment and management decisions within the movie industry.