Mineração de dados na base de dados do jogo league of legends: construção de modelos preditivos de vitórias e derrotas de uma equipe

Abstract

The objective of this work was to build two predictive models to classify match results, the first model predicts matches using data from the first 15 minutes of the match and the second model analyzes the data at the end of the game, based on team performance data in the context of the match. League of Legends game, using database knowledge discovery techniques. Bibliographic research on League of Legends was carried out in order to understand and conceptualize the context of the game and data mining techniques. As for the technical procedures, it is a bibliographical and experimental research. The work proposed the use of Logistic Regression, Random Forest and Decision Tree algorithms. The results obtained met the research objectives, in this way, for the first model the Logistic Regression algorithms of Random Florest had good classification results above 70% of accuracy, however, the decision tree had a result of only 64% of classification. In the second model, all algorithms correctly classified data above 98% accuracy. In addition, this study provided an understanding of the degree of importance of the variables for each model, in addition to the processes of cleaning, transformation, redundancy analysis and selection of variables.