잉글랜드 프리미어리그 공식기록을 통한 경기력의 군집분석

Abstract

The purpose of this study is to identify the performance clustering by ranking and the characteristics of clusters based on the official records of eight seasons from 10/11 to 17/18 of the English Premier League provided by Whoscored. A total of 2,400 official records of 20 teams were collected each season to carry out the study. As the independent variables, the official records for goal, shot, yellow, red card, possession, pass, Aerial Won, Shots Conceded, tackle, interception, foul, offside, Shots On Target, dribble, Foulded were selected, and dependent variables were selected by ranking (A(1-6th), B(7-20th). The statistical package R 3.5.1 was used in conjunction with R studio 1.1.456 for processing the collected data, the K-means cluster analysis was used for clustering the data, and the Silhouette evaluation was used to select an objective optimal initial cluster k. A discriminant analysis was also conducted to identify the relationship between dependent variables and clusters. The results of the study indicated that the initial number of k was selected as 2 and that there was a relationship between the groups according to the ranking(xSUP2/SUP=82.545, p=0), and that ranking A had a higher frequency than that of the group 1 (6.3%), and that ranking B had a higher frequency than that of the group (12 percent). Ranking A and Cluster2 showed relatively high goal, shot, possession, pass, tackle, offside, shot on target, dribble, and foulded than Ranking B and Cluster1 and yellow, red card, aerial won, shot conceded, interception and foul showed relatively low.