Clustering-Based Classification of Student Dropout Patterns: A Case Study in College of Industrial Technology — Misrata,Libya

Abstract

Student dropout remains a critical challenge for educational systems, adversely affecting graduation rates and institutional resource allocation. This study proposes an analytical framework utilizing unsupervised clustering algorithms to classify students at the College of Industrial Technology, Misurata, Libya into two distinct categories: one group comprising students who dropped out early with minimal academic engagement, and another group consisting of students who withdrew after achieving a relative degree of academic progress. This classification aims to provide a deeper understanding of the diverse pathways of student attrition The research integrates demographic variables (e.g., gender, admission age, and enrollment year) with academic performance indicators to construct a comprehensive predictive model. The performance of K-Means and Agglomerative Clustering algorithms was evaluated using validation metrics: Silhouette Score, Davies-Bouldin Index. The findings reveal statistically significant patterns that enable the identification of high-risk student cohorts, providing actionable insights for targeted academic interventions. These results may contribute to the enhancement of student retention policies and support data-driven decision-making.