한국 프로축구 경기력에 따른 특성 군집화

Abstract

This study examines the clustering of performance characteristics in Korean professional football. The study analysed data from three seasons (2019, 2021, and 2022) provided by the K League Portal's data center (data.kleague.com), excluding 2020, which was a reduced season due to COVID-19. The data was standardized using the statistical package R 4.2.3 as a data processing method. The initial number of clusters was selected and graphed based on the Silhouette evaluation criteria to determine the objective initial number of clusters, k. The k-means cluster analysis was performed using the SPSS 27.0 Version statistical package program. Non-parametric statistical techniques, specifically Spearman Correlation Analysis and Frequency Analysis, were employed to determine the association between variables. Additionally, the Mann-Whitney U test, a non-parametric statistical test, was used to compare the differences between each cluster. The results indicate that the official records of the Korean professional soccer K League 1 could be clustered into two distinct groups. Cluster 1, which was dominated by the higher-ranked teams on average among the two clusters, showed significant differences in basic variables such as goals, ball share, and pass success rate, as well as offensive variables such as shots on goal, shots in the penalty area, and shots, and pass types such as pass, long pass, and key pass, compared to cluster 2. On the other hand, Cluster 2, which was dominated by lower-ranked teams on average, only showed a significant difference in aerial contests among the defensive variables.