Genetic Algorithm–Optimized Clustering for University Promotion Target Recommendation

Abstract

Competition among higher education institutions demands promotional strategies that are more targeted and data-driven. This study proposes a clustering-based recommendation model for determining university promotion targets by integrating Genetic Algorithm (GA) optimization into three clustering methods: K-means, Fuzzy C-means (FCM), and K-medoids. The dataset consists of 925 student records (cohorts 2021–2023) from the Information Technology Department, with the selected attributes including school origin, NPSN, school location (city and province), and GPA. Clustering performance was evaluated using the Davis-Bouldin Index (DBI) and the Silhouette Coefficient as primary metrics, with intra- and inter-cluster distances as supporting indicators. The results show that GA-K-means achieves the best performance at K = 3, with a DBI of 1.2792 and a Silhouette Coefficient of 0.2876, and the improvement is statistically significant (p < 0.05). GA optimization also improves FCM performance but does not significantly improve K-medoids performance. Although the GA increases computational time by approximately two to three times, the improvement in clustering quality justifies its use in non-real-time decision-support scenarios. The proposed model enables universities to determine promotion targets in a more objective, adaptive, and data-driven manner, supporting strategic decision-making in higher education promotion.