Medal Prediction Model Based on Machine Learning

This report uses machine learning models to analyze Olympic medal data and uncover hidden insights. We applied KNN to handle missing data and Box-plot to remove outliers. We then built a Mixed Effects Model to identify factors influencing medal counts and used XGBoost to predict the 2028 Olympic medal table, achieving MSEs of 1.0166 (gold), 4.3638 (silver), and 7.8201 (bronze). Additionally, we quantified resource investment using competition and athlete numbers, predicted future trends with ARIMA, and used logistic regression to identify potential first-time medalists like Luxembourg, Angola, and Guinea. We employed the DID method to evaluate the impact of excellent coaches on medal outcomes and used Random Forest to predict the benefits of investing in coaches for specific events, such as France's Basketball, Canada's Water Polo, and the UK's Hockey. We categorized events using HHI and Shannon Diversity Index, finding that evenly distributed events attract more participating countries, promoting sport development. Finally, sensitivity analysis confirmed the XGBoost model's robustness, though some parameter definitions require refinement for improved accuracy.