Players’ Value Prediction Based on Machine Learning Method

Abstract

It is important to predict football players’ value, especially during transfer period. This paper uses the player information and value data of the game FIFA 18 as data source. It is able to realize the prediction of its players’ best positions and values. After reducing the dimensionality of the value prediction model, a cluster analysis on the player’s position is introduced, and then grid search method is adopted to adjust the Xgboost parameters, Finally, Xgboost method is used to predict the player’s worth. The experimental results show that certain accuracy is achieved, but t there is still room for improvement in the accuracy of prediction. Discussions based on experiment results are made.

Measuring the popularity of football players with Google Trends

“…MLR is a classical method that is used as a benchmark, and the RF and GBM methods were selected due to the good results they have achieved in most of the research fields in which they have been applied. In addition, previous researchers have used these machine learning techniques to determine the market value using the RF [10] and gradient boosting, namely XGBoost [39][40][41]. Note that both the RF and GBM make it possible to analyse the importance of a variable.…”