Dozhdikov A.V. Forecasting the Financial Efficiency of Russian Cinema Using a Multifactor Ensemble Machine Learning Model Trained on Historical Data. Finance: Theory and Practice. 2025;29(6):243-268. https: ...

Текст статьи на сайте журнала URL: https://financetp.fa.ru/jour/article/view/4074 (дата обращения 06.01.2026)

Abstract

The object of the study is data on the distribution of Russian cinema films from July 2022 to September 2023. Specifically, it analyzes 185 films that were released during this period. The research tool consists of 26 and 146-factor machine learning models that have been pre-trained based on previous periods (from 2004 to July 2022, with 1,500 films). The purpose of the study is to demonstrate that machine learning models, trained on historical data, can accurately predict future data, which is especially important for funding programs aimed at developing national cinema in the Russian Federation and attracting private investment, in light of the departure of foreign film distributors from the film market. The study used methods to evaluate film projects based on their historical profitability using rental indicators and the characteristics of the creative teams involved in producing them. The emphasis is on ensemble models –AdaBoost, Bagging, ExtraTrees, GradientBoosting, RandomForest, Stacking, Voting, XGBoost, CatBoost. The novelty of this research lies in introducing of new sources into the scientific community and the potential for practical application of the developed methods for both public and private investors to evaluate film projects prior to the start of the production cycle. Conclusions: Conclusions: Based on the analysis of the quality metrics (accuracy, ROC AUC, and others) for a sample of 185 newly released films (through September 2023), we found that the drop in these metrics was not significant. This suggests that it is possible to use pre-trained models based on historical data to make predictions about fees and other rental outcomes. By analyzing the past work of the project director, screenwriters, cameramen, producers, artists, editor, composer and key actors of the project, estimated distribution data, and the amount of project funding, it is possible to make an accurate prediction about the success of a film. This will allow you to see the total fees, payback period, number of views, and viewer rating.