In this study, 42 different types of high schools in Istanbul from which students took University Placement Exam (LYS) are clustered in terms of their performances. It was also aimed to determine the types of tests that are more efficient among these schools. For this purpose, educational data mining techniques such as clustering and decision tree are used. By deploying the non-hierarchical k-means algorithm, schools are separated into 5 different clusters which have different success level for each of Math-Science (MS), Language and Math (LM) and Language-Social Studies (LS) test scores. It is found that Science High Schools, Private Science High Schools, Anatolian High Schools and Anatolian Teacher Schools found to be in the highest achievement level in all of the test scores. Furthermore, constructed decision tree models with CHAID algorithm show that (1) Chemistry for the score type MS, (2) Math for the score type LM and (3) Turkish Language and Literature for the core type LS were the test types which are primarily effective in the division of schools into clusters.

Keywords: LYS (University Placement Exam), educational data mining, cluster analysis, decision tree