Previsão do vencedor do Oscar de melhor filme: uma abordagem com algoritmos de classificação

Abstract

This work presents a quantitative approach to predict the Academy Award for Best Picture winner using supervised machine learning techniques. Based on historical data from major awards such as BAFTA, DGA, PGA, and SAG, covering the years from 2004 to 2023, binary variables were created to indicate each film’s performance during the awards season. Using this structured dataset, Logistic Regression, Decision Tree, and Random Forest algorithms were applied and evaluated through accuracy, precision, recall, and F1- score metrics. In addition, statistical analyses such as correlation, Chi-Square test, and co-occurrence were conducted to identify association patterns between awards. The practical application of the models for predicting the 2025 Oscars showed consistent results, with Random Forest achieving the best performance. The findings suggest that even in subjective contexts such as film awards, it is possible to identify predictive historical patterns, expanding the role of data science in cultural domains.