Research on Olympic Medal Prediction Based on Random Forest Regression

Abstract

The distribution of Olympic medals serves as a critical measure of national sports competitiveness. This study proposes a predictive model based on Random Forest Regression and feature engineering, trained on historical Olympic data (1896-2020) and validated using 2024 Paris Olympics data. Key features, including athlete participation, host country status, historical medal averages, and elite athlete count, were selected through data preprocessing, with model stability assessed via 5-fold cross-validation. Results demonstrate that the Random Forest Regression model achieved a coefficient of determination (R²) of 0.96, with mean squared errors of 0.94 (gold medals) and 3.09 (total medals), outperforming multiple linear regression and support vector regression models. Quantitative analysis reveals a 5%-10% host country advantage in medal probability and a 70% breakthrough potential for non-winning countries through sports diplomacy and elite coaching. These findings provide a data-driven framework for optimizing resource allocation and strategic planning in international sports governance.