Predicting the 2028 Los Angeles Olympic Medal Table: A Machine Learning Approach with Gradient Boosting and Random Forest Models

Abstract

This study predicts gold and total medal counts for the 2028 Los Angeles Summer Olympics, analyzing key factors such as the "star coach" effect and event settings to provide strategic insights for National Olympic Committees (NOCs). Traditional Olympic medal predictions rely on near-event data, with limited consideration of historical trends, while recent machine learning models leverage past data but face challenges with overfitting. Using data from the 1896–2024 Summer Olympics, countries are categorized as traditional, emerging, and potential sports powerhouses. Tailored features are selected, applying Gradient Boosting Regression Tree (GBRT) to the first two categories and Random Forest Regression to the latter. The prediction models are evaluated through R^2, MSE, and MAE, achieving an R^2 of 0.65, MSE of 24.2, and MAE of 24.8. The United States and China are projected to lead with 92 total medals and 39 golds each, followed by Great Britain, Australia, and Japan. This framework offers reliable predictions, highlights the impact of historical trends and coaching influence, and aids NOCs in effective resource allocation and strategy development. The novelty of this approach lies in its combination of historical data with machine learning techniques, offering a more comprehensive method for Olympic medal forecasting.