High-Level Machine Learning Framework for Sports Events Ticket Sales Prediction

Abstract

As all the other live events, sports were particularly affected by the covid-19 pandemic. Nevertheless, with the vaccination campaigns and the expected mass immunization chances are that the social life might soon come back to that was normal before 2020. This paper presents a high-level framework for predicting match ticket sales for a football/soccer club activating in a national championship. Written in R and available as a collection of scripts in a GitHub repository, the framework relies heavily on two R ecosystems of packages for data processing and modeling, tidyverse and tidymodels. For illustration, the framework was applied on a data set provided by a struggling team in Romanian first football league. Predictors relate to expected weather conditions at the start of the game, match day of the week and the starting hour, the phase of the season, and also the team's most recent performances relative to the current match. Despite the dataset limits, results of exploratory data analysis and predictive models are encouraging not only in estimating the match ticket sales, but also in identifying the most important variables associated with ticket sales variability.