A novel machine learning method for estimating football players’ value in the transfer market

Abstract

Every year a huge amount of money is invested by the football clubs in the transfer window period to hire or release
players. Estimating players’ value in the transfer market is a crucial task for the managers of the clubs. Also, it is one of the
attractive aspects of football for fans. Tranfermarkt.com is a reference website that determines the transfer fee of the
players based on its members’ opinions. The limitation of this website has attracted the attention of data scientists in recent
years, resulting in creating datasets and data-driven estimating methods. In this paper, a novel method for estimating the
value of players in the transfer market, based on the FIFA 20 dataset, is proposed. The proposed method has two phases. In
the fi rst phase, the dataset is clustered using an automatic clustering method called APSO-clustering. This automatic
clustering method, which can detect the proper number of clusters, has divided the dataset into 4 clusters automatically
indicating the position of the players: goalkeepers, midfi elders, defenders, and strikers. In the second phase, a hybrid
regression method which is a combination of particle swarm optimization (PSO) and support vector regression (SVR), is
used to build a prediction model for each clusters’ data points. In this hybrid method, PSO is used for feature selection and
parameter tuning of SVR. The achieved results show that the proposed method can estimate the players’ value with an
accuracy of 74%. Comparing the performance of PSO with 3 other metaheuristics, the results demonstrated the superiority
of PSO over GWO, IPO, and WOA.