Search

With the large amounts of structured and unstructured data available to the public, studies on Esports forecasting are increasing day by day. Although prediction studies for esports events are greatly affected by the human factor, it increases the success of predictions with its structure that offers many important parameters in achieving accurate outputs. In this context, it is important how to create models and which machine learning algorithms to choose. In this study, classifications were carried out using various machine learning algorithms to predict the results of the rounds in the online game Counter-Strike: Global Offensive. In the research, a total of seven supervised classification algorithms, namely Logistic Regression, Decision Trees, Random Forest, XGBoost, Naive Bayes, K-Nearest Neighbor and Support Vector Machine were used. Accuracy, Precision, Sensitivity, F-Score and AUC values were calculated in the performance measurement of these algorithms. In addition, algorithms are compared by evaluating ROC curves and confusion matrix. As a result of these measurements and evaluations, the Random Forest algorithm was the most successful algorithm with an accuracy rate of 88%. In addition to these, some suggestions were made for Esports organizations by conducting Exploratory Data Analysis in the context of the winning status of the rounds. [ABSTRACT FROM AUTHOR]