A robust method for clustering football players with mixed attributes
Pierpaolo D’Urso, Livia De Giovanni, Vincenzina Vitale
Abstract
Abstract A robust fuzzy clustering model for mixed data is proposed. For each variable, or attribute, a proper dissimilarity measure is computed and the clustering procedure combines the dissimilarity matrices with weights objectively computed during the optimization process. The weights reflect the relevance of each attribute type in the clustering results. A simulation study and an empirical application to football players data are presented that show the effectiveness of the proposed clustering algorithm in finding clusters that would be hidden unless a multi-attributes approach were used.
Topics & Concepts
Cluster analysisFuzzy clusteringTheory of computationData miningComputer scienceMeasure (data warehouse)Relevance (law)Process (computing)Variable (mathematics)MathematicsArtificial intelligenceAlgorithmLawMathematical analysisOperating systemPolitical scienceData Mining Algorithms and ApplicationsTime Series Analysis and ForecastingData Management and Algorithms