Fecha
2025-10
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editorial
Citas
0 citas en 
0 citas en 
Resumen
El campo del aprendizaje automático y Data Science ha experimentado un crecimiento exponencial en la última década, hasta el punto en el que casi todos los ámbitos y disciplinas han encontrado diversas aplicaciones de estas tecnologías con finalidades de todo tipo. El mundo del fútbol no es un caso aparte y en la actualidad, son numerosos los clubes que poseen un departamento de científicos de datos encargados de buscar soluciones a los distintos problemas a los que pueda enfrentarse un equipo, ya sea mediante el análisis de datos o mediante el desarrollo de modelos. No obstante, existe un gran bache entre los datos a los que puede acceder un usuario normal y aquellos a los que accede un profesional del sector, debido principalmente a la privatización a la que está expuesto el sector, provocando que la mayor parte de fuentes de información acerca de los jugadores, especialmente aquellas relacionadas con datos médicos, sean muy exclusivas y de pago.
La principal tarea que persigue este trabajo es la creación de un dataset desde cero y que sea totalmente replicable por parte de un tercero mediante técnicas de web scraping para posteriormente y una vez creado el dataset, poder desarrollar técnicas de aprendizaje automático para la detección de lesiones en jugadores de fútbol. En concreto, se utilizan tanto algoritmos de clasificación de lesiones, distinguiendo entre aquellas que se consideran de gravedad y las que no, como algoritmos de regresión para tratar de predecir el número de días que estará de baja un jugador.
Para la construcción del dataset se han realizado técnicas de web scraping sobre las web de Transfermarkt y FbRef junto con información de atributos de los jugadores procedentes del videojuego Fifa y disponibles en repositorios en Kaggle. Respecto a la clasificación, tras analizar diversos algoritmos, el algoritmo Naive Bayes es el que mejores resultados presenta, con precisiones de entorno al 59 % y 47 %, dependiendo de la definción de la gravedad de la lesión, mientras que para la regresión no se pudo encontrar ningún modelo que ofreciese un rendimiento aceptable.
The field of Machine Learning and Data Science has experienced exponential growth over the last decade, to the point where almost all areas and disciplines have found various applications for these technologies for all kinds of purposes. The world of soccer is no exception, and today, many clubs have a department of data scientists responsible for finding solutions to the various problems a team may face through data analysis and model development. However, there is a huge gap between the data accessible to a normal user and the data accessible to a professional in the sector, mainly due to the privatization to which the sector is exposed, meaning that most sources of information about players, especially those related to medical data, are very exclusive and require payment. The main task pursued by this work is the creation of a dataset from scratch that can be fully replicated by a third party using web scraping techniques. Once the dataset has been created, machine learning techniques can be developed to detect injuries in soccer players. Specifically, both injury classification algorithms, classifying between those considered serious and those that are not, and regression algorithms are used to try to predict the number of days a player will be out of action. In order to construct the dataset, web scraping techniques were applied to the Transfermarkt and FbRef platforms, complemented with player attribute information derived from the FIFA video game and made available through repositories on Kaggle. With respect to the classification task, after evaluating several algorithms, Naive Bayes yielded the most promising results, achieving accuracies of approximately 59% and 47%, depending on the operational definition of injury severity. Conversely, for the regression task, no model demonstrated a level of performance that could be considered satisfactory.
The field of Machine Learning and Data Science has experienced exponential growth over the last decade, to the point where almost all areas and disciplines have found various applications for these technologies for all kinds of purposes. The world of soccer is no exception, and today, many clubs have a department of data scientists responsible for finding solutions to the various problems a team may face through data analysis and model development. However, there is a huge gap between the data accessible to a normal user and the data accessible to a professional in the sector, mainly due to the privatization to which the sector is exposed, meaning that most sources of information about players, especially those related to medical data, are very exclusive and require payment. The main task pursued by this work is the creation of a dataset from scratch that can be fully replicated by a third party using web scraping techniques. Once the dataset has been created, machine learning techniques can be developed to detect injuries in soccer players. Specifically, both injury classification algorithms, classifying between those considered serious and those that are not, and regression algorithms are used to try to predict the number of days a player will be out of action. In order to construct the dataset, web scraping techniques were applied to the Transfermarkt and FbRef platforms, complemented with player attribute information derived from the FIFA video game and made available through repositories on Kaggle. With respect to the classification task, after evaluating several algorithms, Naive Bayes yielded the most promising results, achieving accuracies of approximately 59% and 47%, depending on the operational definition of injury severity. Conversely, for the regression task, no model demonstrated a level of performance that could be considered satisfactory.
Descripción
Categorías UNESCO
Palabras clave
Fútbol, Aprendizaje Automático, Ciencia de Datos, Clasificación, Regresión, Conjunto de Datos, Scraping Web, Lesiones en el deporte, Football, Machine Learning, Data Science, Classification, Regression, Dataset, Web Scraping, Sports injuries
Citación
Ariza López, Luis. Trabajo Fin de Máster: "Aplicación de técnicas de Aprendizaje Automático para la detección temprana de lesiones en jugadores de fútbol". Universidad Nacional de Educación a Distancia (UNED), 2025
Centro
E.T.S. de Ingeniería Informática