Using machine learning pipeline to predict entry into the attack zone in football
Example Data: https://doi.org/10.6084/m9.figshare.19222746
With the player's position, we create a graph representation, where the vertices represent the players and the edges the possibility to players makes passes between them:
This represetation allows us to exctract metrics from these complex networks, in this work 8 metrics were obtained:
- Betweenness Centrality
- Eccentricity
- Global Efficiency
- Local Efficiency
- Vulnerability
- Clustering Coefficient
- Entropy
- PageRank
So each frame of video was generated a graph with the metrics, the image below demonstrate an example where the size of each node changes conform the player's Betweenness Centrality:
How we analyze a interval of the initial 5 seconds, all the metrics extracted are converted in a visual rhythm image, where in x axis represents the pass of time and y the player's metrics value (high values with light tons and low values closes to black):
In this way is possible to represent the game time series in a unique image:
If you use this for academic research, please cite it using the following BibTeX entry.
@article{stival2023using,
title={Using machine learning pipeline to predict entry into the attack zone in football},
author={Stival, Leandro and Pinto, Allan and Andrade, Felipe dos Santos Pinto de and Santiago, Paulo Roberto Pereira and Biermann, Henrik and Torres, Ricardo da Silva and Dias, Ulisses},
journal={PloS one},
volume={18},
number={1},
pages={e0265372},
year={2023},
publisher={Public Library of Science San Francisco, CA USA}
}