Olympic Medal Prediction Method Based on Random Forest and Multilayer Perceptron Models

Abstract

This paper constructs a medal count prediction model based on historical Olympic data, utilizing the Random Forest algorithm to model nonlinear relationships within complex datasets. First, the raw data undergoes feature selection and correlation analysis to extract key input features from variables such as athlete potential, host country factors, number of events, historical medal records, and gender. The statistical correlations among these features are verified using a correlation matrix. Building on this foundation, multiple training subsets are generated using Bootstrap sampling. A Random Forest prediction model is constructed through a multi-decision tree ensemble mechanism, and model parameters are optimized by adjusting the number of nodes and decision trees. To validate the model’s performance, multi-layer perceptron, multiple linear regression, and support vector regression models are also constructed for comparative experiments. Comprehensive evaluation is conducted using metrics such as mean squared error, root mean square error, mean absolute error, and coefficient of determination. The experimental results indicate that the Random Forest model outperforms the other models in terms of prediction accuracy and stability. It effectively captures the nonlinear characteristics within complex data and achieves accurate predictions of medal distributions, demonstrating good generalization ability and practical value.