This is university project which was worked on by Moritz Kniebel, Vanessa Tsingunidis and myself, in the winter semester 18/19 at the University of Tübingen. As such not all of the work presented here was done by me. I've talked it over with both of them, and they have agreed to me publishing this on my github. In case you want to contact them, please message me. They have stated interest in being publicly related to this project, but I do not wish to publices there contact information on my private github account without their consent.
Most of the text in this project, beside the code of course, was to be in german, as we were to be graded on it. I can provide translations if asked, but as of now none have been necessary and as such none exist.
This project repository was orignally hosted on a private git remote by the mpg, and has been retroactively migrated to github for public viewing access. As such some things needed altering or adding. Things like this section, the licence, and the files for the project presentation have been added since the completion of the project, nothing else has been or will be altered, as this is the state of the project it was graded on. Exceptions to this may include alterations related to security issues.
Konstruktion eines Machine Learning Systems dass die Ergebnisse der Bundesliga eine Woche im Voraus vorhersagt.
Der Ausgang eines Bundesligaspieles hängt von vielen Faktoren ab. Die relative Stärke eines Teams ist nicht über alle Zeit konstant. Die Tagesform der Spieler, Auswärts- und Heimspielsituation, die Jahreszeit, und selbst das Wetter nehmen Einfluss.
Die Teams werden ein System entwickeln das sowohl historische als auch tagesaktuelle Daten aus dem Netz zieht, und daraus Vorhersagen für die Ergebnisse des kommenden Spieltages erstellt. Zu den hierzu nötigen Schritten gehört die Entwicklung einer automatischen Datenabfrage genauso wie die Entwicklung eines prädiktiven Machine Learning Systems.
Mögliche Stretch-Goals sind die Verwendung zusätzlicher Datenquellen (Wetter, Social Media,...), und die Verfeinerung des lernenden Systems von einfachen auf modernere Modelle.