Predicting the Oscars with machine learning

Abstract

Each year before the Academy Awards are announced, several experts try to forecast the winners. I applied statistical learning models to create my own forecast for 2018 in the six main categories, and to identify the most important indicators of a win. These are primarily the results of other award ceremonies. Various kinds of Oscar related data was collected from 1960. Among the three models, the Random Forest Classifier performed best, achieving a 91.5% overall accuracy, and 100% in the 2018 season. Support Vector Machines also performed well overall. Logistic regression was useful to explain the relationship between variables and the nominated films’ chances of victory. For example, the more nominations a film receives, the higher its chance is to win any award.