Development of a Machine Learning Model for Age Prediction of Footballers

Abstract

This study developed a novel age prediction model that can be adopted in sport especially for football athletes to curb some problems  associated with selecting players, and most especially, to minimize age  falsification problems. The study aimed at developing an age  prediction model applying deep learning with machine learning algorithms. The study acquired age dataset from the FIFA website. The  dataset was downloaded in CSV format which contains information about the players, especially their age. Subsequently, we established  a database where each player's image was labelled and mapped with their corresponding CSV file. Euclidean Distance (ED) was used for  feature reduction. The study employed neural networks with deep learning (DNN) and regression using support vectors (SVR) to develop  a model for age prediction. Utilizing Accuracy score, Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Mean Squared  Error (MSE), the effectiveness of support vector regression and deep neural network models was assessed. The model was simulated  using Anaconda Jupyter IDE. The results showed that the Deep Neural Network model has the MSE value of 49.93, RMSE value of 6.92, MAE value of 5.50 and 81% accuracy, while Support Vector Regression has the MSE value of 35.79, RMSE value of 5.98, MAE value of 5.2  and 82% accuracy. The outcome of the age prediction model developed with the DNN and SVR revealed that the SVR model  outperformed DNN. The study recommends that the age prediction model can be used in sport to help the managers in decision making,  especially to minimize age falsification problems.