Spor veri madenciliği tekniklerinin incelenmesi

Sports data mining is the usage of data collected on players, teams, and games for performance evaluation, player selection, score-outcome prediction, and strategy development with data mining tools and techniques. Special performance measures developed for each sports branch have an important role in sports data mining and sports statistics. Performance measures calculated for the team sports and the players can be used to predict the expectation of winning. The Pythagorean Expectation developed for this purpose was first used in baseball games. The Pythagorean Expectation, which attracts the attention of other sports branches, is adapted to other team sports with two possible outcomes such as basketball. However, Pythagorean Expectation studies are limited for sports which have three possible outcomes such as football. In this thesis, it is aimed to investigate sports data mining studies related to various sports branches. In this context, performance measurements and sports data mining studies related to many sports branches are examined and a detailed literature review is performed. Also, it is aimed to suggest a new approach to computation of Pythagorean Expectation for football. In the application section, for the teams competing in fifteen different European football leagues, 2017/2018 season-end rankings and points are predicted using the proposed approach. The data of the last five seasons of the selected European football leagues is used as training data set. All calculations are performed in R.