|
Shape Clustering Based on Spatio-Temporal Data for Analyzing the Collective Behavior of a Football Team
|
Ali Zare Zardiny *   , Zahra Bahramian |
| College of Engineering, University of Tehran & School of Surveying and Geospatial Engineering |
|
|
Abstract: (89 Views) |
In recent years, the analysis of football data with the aim of investigating the behavior of players and matches has received a lot of attention. An important part of these data has a spatio-temporal nature, and this makes the importance of spatio-temporal analyzes more prominent in the football industry. The purpose of this research is to analyze the collective behavior of players at the macro level. For this purpose, in the first step in each time frame, team characteristics are extracted based on a set of spatial, geometric, topological and distribution parameters. Then, these parameters are the basis of team shape clustering. This clustering is performed in two phases. In the first phase, the main clusters are obtained based on the spatial parameters and the defensive or offensive status of the team is determined. In the second phase, for each of the main clusters, based on other descriptors, new sub-clusters are defined. In this research, the data of a football match was used and five main clusters, as well as five sub-clusters for each of the main clusters, were identified. In the evaluation process, the difference between the shape of the team and the center of the corresponding cluster has been measured. The standard deviation of this difference in the main clusters varies between 0.19 and 0.27. Based on this change in the standard deviation, the fluctuations in the team's shape in different areas of the pitch and based on the clustering of the team's overall shape at different times of the match, the time contribution of the clusters and the degree of team dominance on the pitch, the defensive or offensive situation of team and also, overall flow of the team's movement will be determined. Considering the spatial, topological and density parameters along with the geometrical parameters, performing clustering in two stages and without the need to transfer the shape to the raster space (and as a result no need to use image processing techniques) are the most important distinguishing points of the proposed method compared to previous research.
|
|
| Keywords: Analysis of collective behavior of football players, spatio-temporal data, convex hull, two-phase clustering, K-Means algorithm |
|
|
|
Type of Study: Research |
Subject:
GIS
Received: 2024/07/17 | Accepted: 2024/10/8 | ePublished ahead of print: 2024/10/29
|
|
|
|
|
|
|
| Send email to the article author |
|
|
