Modeling of placement situations in secondary education institutions according to LGS preferences with data mining approach

This research aims to model the placement of students in high schools each year based on their preferences in the LGS using DM methods. The study group consists of 6000 8th-grade students who made their LGS preferences in Ankara province in 2022. Following the LGS results, students are placed in schools with exams and without exams, while some students fail to be placed in any school due to various reasons. To enable algorithms to conduct better testing and prediction, 2000 records were randomly selected based on their placement status: those who were unable to be placed in any school, those placed in a school with exams, and those placed in a school without exams. Orange data mining software was used for data analysis. The research selected the top 10 data headings from the dataset using the information gain method, and attempts were made to model students' placement statuses (unplaced, placed in a school with exams, placed in a school without exams) using DM methods based on this selection. It was found that end-of-year academic success scores were more effective in indicating placement, while other demographic data such as parents' education level, students' second and third preferences, number of preferences, whether they receive scholarships, and the type of school they are enrolled in (public middle school, private middle school, religious middle school) followed in importance. When examined in terms of classification algorithms, logistic regression stands out with the highest accuracy rates at 67%, followed by artificial neural networks and naive Bayes algorithms at 64%. As a result of this study, an application similar to a preference robot could be offered to students based on their academic and demographic data, and different academic and demographic data headings could be used for further EDM applications.