This work wants to create a small framework of tools which permits the coaches to base their new strategies not only on their intuitions and judgments but also on comprensive (and maybe not human-eyevisible) statistics. Using the past researches in the football and sport elds, it will analyze a match recorded by DEBS with an innovative system of sensors which could be the future of football.
The full dataset can be retrived here: http://www.orgs.ttu.edu/debs2013/index.php?goto=cfchallengedetails but we also split it in a 5m dataset, placed in the project. However for the passage patterns and similar players, our advice is to use the full dataset.
The whole code is written in Python, so it is multi-platform. The main.py file is the analyzator of the dataset, that need to be placed in the same directory with name "full-game". After the analyzation everything will be saved in a database and you can use kmeans.py for the k-means clustering, hierarchical.py for the agglomerative clustering and edge.py for the passage patterns.
The dataset must be placed in the "DEBSDATA" directory with the name "full-game". To start the project just run "TrajClustering_SpeedPerform.py"
You need thhese libraries:
numpy string select sys matplotlib.pyplot subprocess uuidthe datetime csv math copy thread pylab exceptions
through the stdin you can chose the several options to discover the functionality.
The part of trajectory clustering needs a c program, there already esist the compiled version of a program in /movebank/bin/ (traclus). It is compiled on linux Ubuntu 64 bit. If it should not work recompile it. You will find the make file in the source code directory /movebank/bin/traclus.
For every doubt or comments michele.linardi@alice.it/me@marcodena.it