Title: Machine learning approach to predicting a basketball game outcome

Authors: Roger Poch Alonso; Marina Bagić Babac

Addresses: Barcelona School of Informatics, Polytechnic University of Catalonia, C/Jordi Girona, 1-3 Edifici B6 Campus Nord, 08034, Barcelona ' Faculty of Electrical Engineering and Computing, University of Zagreb, Unska 3, HR-10000, Zagreb, Croatia

Abstract: The outcome of a basketball match depends on many factors, such as the morale of a team or a player, skills, coaching strategy, and many others. Thus, it is a challenging task to predict the exact results of individual matches. This paper shows how to learn from historical data about previous basketball games, including both individual and team features, to predict future matches. It outlines the advantages and disadvantages of existing machine learning systems and tries to apply the best practices focusing on a case study of the National Basketball Association (NBA). In addition, a comparison between different machine learning algorithms in search of the most accurate prediction is provided.

Keywords: machine learning; supervised learning; prediction; KNN; k-nearest neighbours; decision trees; Naive Bayes classifier; basketball; NBA.

DOI: 10.1504/IJDS.2022.124356

International Journal of Data Science, 2022 Vol.7 No.1, pp.60 - 77

Received: 29 Sep 2021
Accepted: 14 Jan 2022
Published online: 25 Jul 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article