Prediction and Strategic Analysis of Olympic Medals

Abstract

The Olympic medal table showcases the competitive level of each country, and establishing a medal prediction model based on historical data is crucial for accurately predicting each country's performance at the 2028 Los Angeles Summer Olympics. This article first preprocesses data from the past 10 Olympic Games, selecting 11 features Xi (such as total number of participants, total number of events, etc.) and 4 predictor variables Yi (such as whether medals were won, total number of medals, etc.), and encodes the classification features. Subsequently, five cross validation fits were performed on the data using random forest, LightGBM, and XGBoost models to establish the relationship between features and predictor variables. Next, the stacked model is used to integrate the prediction results of these three models, assign different weights, and finally construct an Olympic medal prediction model. On the training set (80%), the model scored 0.823, and on the testing set (20%), the model scored 0.806. Next, use ARIMA and grey prediction models to predict the feature variable Xi of the 2028 Olympics, and substitute it into the medal prediction model to obtain the medal situation Yi of each country in 2028. According to the model, China and the United States may perform better in 2028, while Japan and France may perform worse; The probabilities of Monaco, Bahamas, and Brunei winning medals for the first time are 83%, 81%, and 76%, respectively; Diving, weightlifting, shooting, and gymnastics will have a significant impact on China's total medal count; As the host country, there was a significant change in the medal count of Chinese gymnastics in 2008.