Predicting individual event attendance with machine learning: a ‘step-forward’ approach

ABSTRACT

Accurately predicting attendance at live events has important operational and financial implications for the arts, entertainment and sport industries. Advances in machine learning offer the potential to improve these processes. Using 10 rounds of attendance data from 5,946 season ticket holders of one professional football team (i.e. 59,460 decisions), we assess the ability of four machine learning approaches to predict attendance. Our results indicate that two machine learning algorithms, XGBoost and Support Vector Machine (SVM), outperform the most commonly employed methodology for modelling individual sport attendance (i.e. logistic regression), in terms of accuracy, recall, F-score and area under the curve (AUC). Random forest and boosted aggregation (bagging) approaches are also compared. Our results suggest that adopting machine learning methodologies, and in particular, XGBoost and SVM, offers providers of live events an improved ability to understand and predict individual attendance, and insight into which consumers are most receptive to changing attendance decisions.

KEYWORDS:

• event attendance
• machine learning
• random forest
• sport
• XGBoost

JEL CLASSIFICATION:

• C33
• C53
• Z29

Disclosure statement

No potential conflict of interest was reported by the authors.