Comparative Analysis of K-Means and K-Medoids Algorithms in New Student Admission

Abstract

Universitas ‘Aisyiyah Yogyakarta is one of the private universities in Yogyakarta. The large number of private universities in Yogyakarta has intensified the competition for new student admissions. In this situation, every university requires the right strategy to attract prospective students. One of the strategies used by Universitas ‘Aisyiyah Yogyakarta to capture the interest of potential students is by conducting direct promotions to schools in Yogyakarta, Java, and Sumatra. In the admission process for new students in the Information Technology Study Program, a common problem arises, which is the number of prospective students who do not complete re-registration each year. These students pass the selection and are declared accepted, but they do not proceed with re-registration. The school presentation strategy contributes to student admissions, making it a good strategy, but it requires significant operational costs. Promotion area segmentation is needed so that this strategy can be more targeted, resulting in more efficient spending. Segmenting or grouping promotion areas can be addressed using data mining techniques, specifically clustering. This study aims to segment promotion areas using clustering algorithms, namely K-Means and K-Medoids, along with the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. The evaluation of DBI (Davies-Bouldin Index) showed that the K-Means algorithm performed better than the K-Medoids algorithm. The comparison between the K-Means and K-Medoids algorithms was assessed based on the DBI evaluation results, with the smallest DBI value observed in the K-Means algorithm. The DBI value for K-Medoids was 0.196, while for K-Means it was 0.170.