Research on Olympic Medal Prediction Method Based on Ensemble Learning and Grey Prediction

Abstract

This study focuses on predicting Olympic medal counts by integrating machine learning algorithms and grey prediction models. By analyzing historical Olympic data from 1988 to 2024, the study incorporates key factors such as the number of athletes, gender distribution, and host country effects to improve medal forecasts. The research employs Random Forest, XGBoost, and LightGBM algorithms to construct individual models, and then utilizes a stacking approach to combine these models, enhancing prediction accuracy and stability. Each model is trained using historical data, and the prediction results are evaluated through cross-validation to assess generalizability. Furthermore, grey prediction models are used to address data uncertainty and offer supplementary insights into future outcomes. The combined approach demonstrates improved accuracy over individual models, providing a more reliable forecast for Olympic medal outcomes. The results also highlight the significant influence of host country effects on medal performance, with specific recommendations for future competition strategies. This research contributes to the field of sports analytics by combining diverse predictive techniques, offering a robust framework for forecasting Olympic medal distributions, and assisting in strategic planning for upcoming Olympic Games.