Finding Roles of Players in Football Using Automatic Particle Swarm Optimization-Clustering Algorithm

Area of Science:

• Sports Analytics
• Data Science
• Computational Intelligence

Background:

• Professional sports organizations increasingly use data analysis for performance evaluation.
• Analyzing complex team sports like football presents significant data challenges.
• Identifying player roles is crucial for team strategy and performance comparison.

Purpose of the Study:

• To propose an automatic big data clustering method for analyzing football player performance.
• To extract distinct player roles using a swarm intelligence algorithm.
• To enhance team performance analysis and facilitate meaningful player comparisons.

Main Methods:

• Developed an automatic big data clustering method based on a swarm intelligence algorithm (Particle Swarm Optimization).
• The algorithm operates in two phases: determining the number of clusters and finding centroid positions.
• Tested on synthetic datasets and a large real-world dataset (93,000 objects from ~4900 matches).

Main Results:

• The proposed algorithm effectively clusters player performance data.
• Demonstrated superior performance compared to two conventional clustering methods on synthetic data.
• Successfully applied to a large dataset of European football league matches to identify player roles.

Conclusions:

• The swarm intelligence-based clustering method provides an effective approach for automatic player role extraction in football.
• This method can significantly aid in analyzing team dynamics and individual player contributions.
• The approach is scalable and robust for handling large-scale sports performance datasets.