Predicting Teamfight Tactics Results with Machine Learning Techniques
2020 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE credits
Student thesis
Abstract [en]
The scope of this thesis is to determine whether Machine Learning techniques used in predictingtraditional sports results are relevant in predicting electronic sports results. Moreover, it exploreswhether there is a hidden pattern in a team that leads to a win rather than just plain team composition.The selected video game from electronic sports was Teamfight Tactics. The literature study wasconducted on journals, articles and projects that predicted traditional sports outcomes, from which theMachine Learning models were collected. Literature concerning Machine Learning algorithms, theiradvantages and caveats was studied as well for the scientific background and implementation. Finallythe experiment consisted in implementing models identified in literature study and analyzing the results.The experiment showed that predicting electronic sports with Machine Learning techniques used intraditional sports is not only relevant but also more accurate, due to the complex nature of the videogame and data availability which yields a dataset with high dimensionality and data points.The secondpart of the experiment showed that indeed there is a pattern in a team that leads to higher winningchances. It was demonstrated by encoding the dataset with word2vec (encoder used in naturallanguage processing) and visualizing the data through the Embedding Projector (a tool that showspatterns in data through clusters)
Place, publisher, year, edition, pages
2020. , p. 62
Keywords [en]
Machine learning, feature engineering, classification algorithms, regression algorithms, electronic sports results prediction
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:hkr:diva-21158OAI: oai:DiVA.org:hkr-21158DiVA, id: diva2:1463305
Educational program
Bachelor programme in Computer Software Development
Presentation
2020-08-25, Kristianstad, 13:00 (English)
Supervisors
Examiners
2020-09-032020-09-012020-09-03Bibliographically approved