Özet:

Football is one of the most followed sports in the world and in Turkey. This state of popularity of football is used in information technologies and with the developing data science, match statistics can be easily identified. The most important thing about football is the match. Many different criteria affecting the match result (the number of goals, the number of cards the team receives, the weather, playing on the move, etc.) there is. This study used the data obtained from the matches played by the Turkish Football Federation Super League in the 2019-2020 and 2020-2021 seasons. The team’s win and losing conditions are classified and modeled by decision-making methods, which are the main objective of the work. The red or yellow cards received by the host and the rival team in the games are determined as independent variables by categorizing the number of foreign players in the teams and the number of goals scored. Depending on these variables, the winning or losing status of the host team is modeled using the Logistic Regression and Decision Tree (CART, QUEST and CHAID) algorithms. There are six distinct models in the work. Comparing the accuracy percentages, sensitivity, selectivity and F-score values of the created models, it is determined that the best model is the CART algorithm with a accuracy percentage of 67.6% from the decision trees. In this model, the rival red card status and the offensive and defense forces are important in the team's victory or loss. It is also shown that machine learning algorithms can be used in the modeling of football data.

Anahtar Kelimeler: