Bachelor Thesis 2016-17: Analyzing the Goal Contribution of English Club Midfielders

Football is the most popular sport, followed by all the countries. Professional clubs spend huge amounts of money to fetch good players for competition at the highest domestic levels. In its efforts to achieve performance-related goals, clubs have to ensure that they identify the right players. In this research, a data mining technique is used to predict the goal contribution of midfielders from English football clubs, which assists the club in making player selection. With the help of five football seasons (English Clubs: 2011-12 to 2015-16) data, an attempt has been made to identify the potential of midfielders by using a classification model based on characteristics like weather, ground, competition level, timing of matches, opposition strength, substitution status, player form, and position of players. In data mining, various techniques were tested, such as BayesNet, Naive Bayes, Naive Bayes Multinomial, Logistic Regression (multinomial), Multilayer Perceptron, Random Forest, J48, and SMO. Out of these, tool SMO, which solves the quadratic programming problem related to Support Vector Machines (SVMs), gave the best result. The performances were mainly assessed with a confusion matrix that offered information about the prediction of performance by midfielders and the events that included them. In the case of this experiment, SMO was seen to be better off than other algorithms and therefore most suitable as a classifier. Thus, this work focuses on presenting the benefits of data mining approaches, with consideration given to Sequential minimal optimization (SMO) in identifying player performance and advancing the area of sports analytics studies.