Scholarlink

Football generates vast amounts of data daily on players, teams, and leagues, used for performance analysis. However, advanced analysis technologies are often inaccessible to lower-budget teams. This study developed a predictive tool to estimate the number of goals a team might score in a match, focusing on the UEFA Champions League with data from 2005 to 2022, totaling 4237 records. The model development included three stages: database creation, data preprocessing and analysis, and model construction. Data from the Fbref and FIFA Index, including player data from the EA FC video game, were collected using web scraping to build an expandable database. Additional variables, such as team origin league and momentum (average goals in the last five matches), were added to improve model accuracy. The model, a deep fully connected neural network evolved from a multi-layer perceptron, achieved 57% accuracy. The findings suggest that effective analysis tools can be created using data mining and machine learning, supporting lower-budget teams by providing more information and reducing competitive disparities.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Design and Implementation of a Machine Learning Model for Soccer Match Prediction Based on Player Statistics

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Antonio Muñoz Barrientos,
  • Humberto Muñoz Bautista,
  • Alejandro Padilla Díaz,
  • Francisco Javier Álvarez Rodríguez

摘要

Football generates vast amounts of data daily on players, teams, and leagues, used for performance analysis. However, advanced analysis technologies are often inaccessible to lower-budget teams. This study developed a predictive tool to estimate the number of goals a team might score in a match, focusing on the UEFA Champions League with data from 2005 to 2022, totaling 4237 records. The model development included three stages: database creation, data preprocessing and analysis, and model construction. Data from the Fbref and FIFA Index, including player data from the EA FC video game, were collected using web scraping to build an expandable database. Additional variables, such as team origin league and momentum (average goals in the last five matches), were added to improve model accuracy. The model, a deep fully connected neural network evolved from a multi-layer perceptron, achieved 57% accuracy. The findings suggest that effective analysis tools can be created using data mining and machine learning, supporting lower-budget teams by providing more information and reducing competitive disparities.