Title of the thesis a data mining study on the functionality of the criteria scores used in student admission to graduate programs

This study aims to examine to what extent the criteria of ALES (Academic Staff and Graduate Education Entrance Examination), foreign language proficiency, graduate entrance exam and undergradute GPA, widely used in the admission of students to graduate programs, allow an effective evaluation process. The study investigates whether the criterion scores of the students admitted to the graduate programs are an effective predictor of their graduation from the program. Classifications (Prediction) algorithms for data mining such as Logistic regression, Decision trees (C4.5), K-nearest neighbors (KNN) were run on the aforementioned scores, with the expectation that these algorithms would predict whether students would complete the programs they were admitted to. The analyzes were carried out through the Weka 3.9.4 software. As the data set, the study uses ALES, foreign languance proficiency, Graduate entrance exam, and GPA scores of 740 students admitted in non-thesis master's, master's and doctoral programs at Bolu Abant İzzet Baysal University, Institute of Educational Sciences between 2011-2016, as well as their graduation status. While the data from 66% of the 740 individuals in the data set were used by algorithms as learning data to create models that explain the graduation status, and the data from the remaining 33% were used to test the models created. The findings obtained from the study shows that none of the three classification algorithms managed to create a model that could explain the graduation status of students from the programs by using one or several of ALES, foreign language proficiency, GPA and graduate entrance exam scores. This finding leads us to conclude that the criterion scores used in the admission of students to graduate programs fail to predict the student graduate during the program process.