Clustering functional data streams: Unsupervised classification of soccer top players based on Google trends

Abstract

Nowadays, football teams have become large companies, which produce very high incomes, and the induced gain that may arise from purchasing notorious players has become a crucial aspect for their commercial strategies. The interest for a player is not only exclusively tied to his technical skills but also to other factors, which may attract people not only interested in football. Because soccer has become a fact of life for many supporters, the attention paid to football players is also reflected in the internet traffic. For this reason, the analysis of web data nowadays is a central issue for monitoring players' notoriety and catching information about marketing strategies. This paper aims to provide an original methodological approach for the analysis of web data using the functional data analysis approach. The main advantages of the proposal are the drastic dimensionality reduction of the data and the use of functional tools for obtaining additional information about data. Specifically, in this study, we focus on the problem of clustering data streams using the k‐means algorithm and 3 different semimetrics based on the functional principal components decomposition, first and second derivatives, respectively. An application to a real data set regarding Google queries of 24 football players is proposed for illustrating the method. The final aim of this research is provide scholars, practitioners, football clubs, and agencies interested in planning marketing strategies with additional tools for analyzing web data.

Assessing Social Interest in Burnout Using Google Trends Data

“…Since Google Trends data continuously flow from the server of a web site, they can be seen as functions in a continuous domain, rather than scalar vectors (Fortuna et al 2018). Despite the continuous nature of functional data, in real applications, sample curves are observed with error in a discrete set of sampling points, t…”

“…Indeed, people reveal information about their needs, wants, interests, moods and phycological problems through their Internet search histories, which are stored in the form of Google Trends data (Zeynalov 2017). More specifically, we propose to analyze Google Trends data through the functional data analysis (FDA) approach (Ramsay and Silverman 2005;Ferraty and Vieu 2006) because data flowing from the web can be viewed as an infinite process, which continuously evolve over the time domain (Fortuna et al 2018). Since functional data are infinite-dimensional objects, they provide a more suitable representation of Google Trends search queries than traditional multivariate vectors.…”

A robust method for clustering football players with mixed attributes

“…Adhikari et al (2020) propose a methodology for cricket player selection based on an efficiency data envelopment analysis, semi-variance approach, and Shannon-entropy. Cea et al (2020) analyze the procedure used by FIFA up until 2018 to rank national football teams Papers specifically on clustering of sports data are Gates et al (2017), Behravan and Razavi (2021), Fortuna et al (2018), Lu and Tan (2003), Narizuka and Yamazaki (2019), Narizuka and Yamazaki (2020), Shelly et al (2020), Ulas (2021), most of which with applications to football data. Gates et al (2017) propose a unsupervised classification method that defines subgroups of individuals that have similar dynamic models.…”

Uniform Manifold Approximation and Projection Analysis of Soccer Players

“…The data from the time motion and the body acceleration/deceleration features were processed using repeated-measures factorial ANOVA and two-step cluster analysis to classify players. Fortuna et al [ 61 ] analyzed the notoriety and international popularity of players in the viewpoint of Google queries over time. The data streams were processed through K-means clustering and three semi-metrics using the functional principal component decomposition and their first and second derivatives.…”