Publication: Predicción de cambios en el rendimiento de futbolistas
Loading...
Official URL
Full text at PDC
Publication Date
2019
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Nuestro proyecto trata de predecir la mejora de rendimiento entre dos temporadas consecutivas de un jugador de fútbol a través del análisis de datos. El principal objetivo, es ser capaces de determinar si el jugador mejorará su rendimiento en al menos un tanto por ciento dado, y tener además alguna indicación de lo probable que resulta que esto suceda. Para esto último, proponemos utilizar métodos de aprendizaje automático como la regresión logística, que indican qué probabilidad hay de que el elemento sea clasificado en una clase u en otra. Esta probabilidad se verá afectada, a la hora de hablar de su fiabilidad, por la precisión del propio modelo, el cual generaremos a partir de un conjunto de datos de jugadores cedido por la empresa Driblab, que se dedica al recuitring de jugadores y que ha mostrado su interés en este proyecto.
El trabajo trata el preprocesamiento de los datos, la obtención de los ficheros con los que entrenamos y probamos el modelo, los clasificadores que hemos utilizado, y los motivos que nos encaminan a tomar ciertas decisiones.
Cabe destacar que, a pesar de la complejidad del problema, los resultados obtenidos en clasificación son buenos y permiten detectar a jugadores que van a incrementar su rendimiento de forma notable la siguiente temporada.
Our project tries to predict the improvement of performance between two consecutive seasons of a soccer player through data analysis. The main objective, is to be able to determine if the player will improve his performance in at least a given percentage, and also have some indication of how likely it is that this happens. For this reason, we propose to use machine learning methods such as logistic regression, which indicate the probability that the item will be classified in one class or another. This probability will be affected, when talking about its reliability, by the accuracy of the model itself, which we will generate from a set of player data provided by the company Driblab, which is dedicated to recuitring players and has shown interest in this project. The paper discusses preprocessing of data, obtaining the files with which we train and test the model, the classifiers we used, and the reasons that lead us to make certain decisions. It should be noted that despite the complexity of the problem, the results obtained in classification are good and allow detecting players who will increase their performance significantly next season.
Our project tries to predict the improvement of performance between two consecutive seasons of a soccer player through data analysis. The main objective, is to be able to determine if the player will improve his performance in at least a given percentage, and also have some indication of how likely it is that this happens. For this reason, we propose to use machine learning methods such as logistic regression, which indicate the probability that the item will be classified in one class or another. This probability will be affected, when talking about its reliability, by the accuracy of the model itself, which we will generate from a set of player data provided by the company Driblab, which is dedicated to recuitring players and has shown interest in this project. The paper discusses preprocessing of data, obtaining the files with which we train and test the model, the classifiers we used, and the reasons that lead us to make certain decisions. It should be noted that despite the complexity of the problem, the results obtained in classification are good and allow detecting players who will increase their performance significantly next season.
Description
Trabajo de Fin de Grado, Universidad Complutense, Facultad de Informática, Departamento de Sistemas Informáticos y Computación, Curso 2018/2019