Medal Prediction Model Based on BP Neural Network

Abstract

This study predicts medal standings for the 2028 Olympics using a two-layer BP neural network model, analyzing historical data and national potential factors. A newly proposed National Potential Index (P) serves as a key parameter to assess countries' likelihood of winning their first medal. Findings suggest Czechoslovakia, Yugoslavia, Australasia, and Uruguay will achieve medal breakthroughs, while China and the United States maintain dominant positions. The research employs a hierarchical analytical approach to enhance prediction reliability, evaluating model performance through precision, recall, F1 score, MSE, and MAE metrics. Results demonstrate the model's effectiveness in forecasting medal distributions, particularly in identifying emerging competitors and confirming traditional powerhouses' advantages. Validation confirms the model's reference value for Olympic medal predictions, successfully integrating classification and regression analyses to balance breakthrough detection with ranking accuracy. The methodology emphasizes stepwise predictions, first determining medal qualification probabilities before estimating specific medal counts, thereby optimizing outcome reliability. This dual-layer structure addresses both categorical (medal attainment) and continuous (medal quantity) prediction challenges, offering comprehensive insights into Olympic performance dynamics.