Football Player Value Prediction: Comparing Machine Learning Models with Cross-Validation

Abstract

The project mainly focuses on the evaluation of different models used in order to predict the current value of football players based on a dataset from Kaggle. Three models-random forest, gradient boosting, and ridge regression-are being examined using key indicators such as R-squared (R2), root mean square error (RMSE), and mean absolute percentage error (MAPE). A cross-validation process is applied to ensure the robustness of the model evaluations. Among these models, the gradient boosting model is the most suitable since it provides the lowest RMSE and the highest R2 indicating high accuracy. A young, robust, and healthy player who has had a high market value in the past or present is more likely to have a high current value. The research aims to provide team managers with a relatively accurate model to predict player’s market value. At the same time, this study can help players understand which factors impact the most in terms of current value, encouraging them to improve themselves in certain areas.