Analisis Prediktif Performa Pemain Piala Dunia FIFA Menggunakan XGBoost untuk Mendukung Pengambilan Keputusan Pelatih

Abstract

Machine learning allows player performance analysis based on data to improve the credibility of coaches' decision-making. This study aims to create a player performance prediction model for the FIFA World Cup using data containing 54,600 observations and 75 attributes, covering technical, tactical, and physical statistics, as well as indicators of player performance. Data preprocessing, feature selection, splitting into training and testing sets (80:20), and model building were done using Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and XGBoost. The results show that Gradient Boosting delivered the best performance with an MAE of 1.970177, RMSE of 2, and Coefficient of Determination (R2). In the XGBoost model, Feature Importance and SHAP analysis found that playing minutes, tournament rating, passing accuracy, and expected goals (xG) are the main components affecting player performance. This model can generate player rankings, recommendations for the Starting XI players, and substitutes, and it works as a Decision Support System (DSS). The DSS helps coaches make strategies, assess player performance, and decide the team lineup with accurate data.