Using social network analysis and gradient boosting to develop a soccer win–lose prediction model

We present the conceptual framework of a soccer winlose prediction system (SWLPS) focused on passing distribution data (which is a representative characteristic of soccer) using social network analysis (SNA) and gradient boosting (GB). The general purpose of soccer predictions is to help the field supervisor design a strategy to win subsequent games using the derived information to improve and expand the coaching process. To implement and evaluate the proposed SWLPS, actual network indicators and predicted network indicators are generated using passing distribution data and SNA. The winlose prediction is conducted using the GB machine learning technique. The performance of the SWLPS is analyzed through comparison with various machine learning techniques (i.e., support vector machine (SVM), neural network (NN), decision tree (DT), case-based reasoning (CBR), and logistic regression (LR)). The experimental results and analyses demonstrate that the network indicators generated through SNA can represent soccer team performance and that an accurate winlose prediction system can be developed using GB technique. This study proposes a conceptual framework for soccer winlose prediction system.The proposed predicting system employs a social network analysis to generate input variables.A gradient boosting is utilized to simulate predictions based on network indicators.Application to the Champions League is used to validate the proposed system.