The Use of Predictive Methods and Text Analysis for the Purposes of Predicting the Outcomes of Tennis Matches and Developing Betting Strategies

U ovom radu razvijen je sustav za predviđanje ishoda teniskih mečeva kroz upotrebu strojnog učenja i Markovljevih lanaca. U prvim poglavljima objašnjene su osnove tenisa te je opisan korišteni skup podataka i razvijeni prediktivni modeli. Različite varijante razvijenih modela evaluirane su na različitim skupovima značajki, a obavljena je i analiza intervjua igrača koristeći metode NLP-a te su dobivene značajke korištene u modelima. Uz to, rad uključuje izradu i evaluaciju raznih strategija klađenja koje se temelje na dobivenim prediktivnim modelima.In this study, a system for predicting the outcomes of tennis matches using machine learning and Markov chains was developed. The initial chapters explain the basics of tennis and describe the used data set and developed predictive models. Various variations of the developed models were evaluated on different feature sets. Also, player interviews were analyzed using NLP methods and obtained features were incorporated into models. In addition, the study involves the creation and evaluation of various betting strategies based on the obtained predictive models