A Study on Predicting the Medal Situation of the 2028 Olympic Games Based on Machine Learning and Predictive Modeling

Abstract

This paper predicts the medal situation of each country in the 2028 Olympic Games by analyzing data from the past 10 Olympic Games. It first selects 11 features (such as the total number of participants, total number of sports categories.) and 4 predictor variables (such as whether medals were won), and encodes the classification features. Random forest, LightGBM, and XGBoost models are then used to fit the data with five-fold cross-validation, establishing the relationship between features and predictor variables. A stacking model is used to integrate the prediction results of these three models, assigning different weights. The model achieved a score of 0.823 on the training set (80%) and 0.806 on the testing set (20%). Subsequently, ARIMA and grey prediction models are applied to predict the feature variables for the 2028 Olympics, and these are substituted into the medal prediction model to obtain the medal situation for each country in 2028. The results suggest that China and the United States are likely to perform better, while Japan and France may perform worse. Additionally, the paper explores the "Great Coach Effect," analyzing Lang Ping's impact as coach of the 2008 U.S. women’s volleyball team and the 2016 Chinese women’s volleyball team.