Data Mining for Football Performance Analysis and Prediction: A Systematic Literature Review

Abstract

Data mining has become an important approach to analyzing football-related data to support performance evaluation and decision-making. However, the application of data mining techniques in football covers various analytical objectives, ranging from match outcome prediction to player performance analysis and physical load monitoring. This study aims to systematically review the application of data mining techniques in football, identify the methods used, and synthesize the main findings of previous studies. A Systematic Literature Review (SLR) was conducted following the guidelines proposed by Kitchenham. The literature search covered four databases: Google Scholar, ScienceDirect, IEEE, and SpringerLink. Studies were selected based on predefined inclusion and exclusion criteria, resulting in 15 articles published between 2003 and 2023. The reviewed studies were examined in terms of their analytical objectives, data mining techniques, and reported findings. The results indicate that the reviewed literature addresses four analytical categories: estimation, forecasting, classification, and clustering. The techniques discussed include neural networks, linear regression, support vector machines, decision trees, k-nearest neighbors, random forests, and K-means clustering. These techniques have been applied to different football-related tasks, including match outcome prediction, attacking effectiveness analysis, player performance assessment, and training load monitoring. The findings highlight the diverse applications of data mining in football analytics while indicating that reported model performance must be interpreted in the context of each study's dataset, prediction target, and evaluation procedure. This review provides an overview of existing applications and identifies the need for more consistent methodological reporting and comparative synthesis in future research.