Predikce výhry profesionálních zápasů v Dota 2

This thesis is dedicated to the subject matter of win prediction in an esport title Dota 2. The aim of the thesis is construction of models for prediction with the use of data, which are available before the beginning of the match in the shape of selected heroes and data obtained through the progression of a match at the 10 minute mark, such as difference in gold between teams, difference in taken towers or amount of activated runes. There is description of data in the thesis, the construction of logistic regression, random forest and extreme gradient boosting models. The logistic regression model is constructed with all variables except the difference in gold and experience, which have high multicollinearity with the other variables, but are kept in the random forest and extreme gradient boosting models to capture more complex relationships between variables. All the created models are then evaluated with the use of evaluation metrics and their quality of prediction is compared between them.