Prediction Of Medals Based on Machine Learning and OLS Statistical Regression Models

Abstract

This paper constructs a predictive model that integrates variables such as the number of athletes and the growth rate of medals. It employs regression and classification techniques from the field of machine learning, utilizing algorithms including Random Forests, Support Vector Machines, and Logistic Regression and the performance of random forest in the last three Olympic Games was compared to determine the optimal performance of random forest in the classification task. Additionally, it applies Bayesian Change Point Detection and Ordinary Least Squares statistical regression models to analyze the performance trajectories of the Chinese women’s volleyball team, the United States women's volleyball team, and the Romanian gymnastics team. Based on the data predicted by the model, the study proposes to increase investment in sports science research, attract talent, strengthen logistical support, and develop special policies. At the same time, error analysis and sensitivity analysis were carried out in this study, and the limitations in the data and model and the sensitivity of the model to certain features were pointed out.