Olympic Medal Count Prediction Model Based on Xgboost

Abstract

Accurately predicting Olympic medal counts for each country is of significant value for shaping sports policy, optimizing resource allocation, and advancing the sports industry. This study presents a predictive model based on the Extreme Gradient Boosting (XGBoost) algorithm to estimate the number of gold, silver, and bronze medals awarded to each country in the Olympics. The validity of the model is demonstrated by predicting the medal counts for the 2024 Olympics. Relevant data from 1896 to 2024 were collected and preprocessed. Five key features were extracted for each Olympic edition: the host country, the number of gold, silver, and bronze medals, the total medal count, the number of athletes per country, the gender distribution of athletes, and the sports events in which each country participated. Data from 1896 to 2012 were used for training, while data from the 2016 and 2020 Olympics were used as the test set. The model was trained using the XGBoost algorithm, and optimization was performed by minimizing the root mean square error (RMSE). Four features for the 2024 Olympics—host country, number of participating athletes, gender distribution of athletes, and the sports events contested—were used to predict the medal counts for each country. By comparing the predicted results with the actual data, the RMSE values for gold, silver, bronze, and total medals were calculated to be 0.6462, 0.5547, 0.2965, and 0.3922, respectively. These results validate the exceptional performance of the XGBoost model in predicting Olympic medal counts, providing effective and forward-looking strategic insights for the optimization of sports resource allocation and the setting of competitive goals across countries.