PREDIKSI RISIKO DROP OUT MAHASISWA MENGGUNAKAN MODEL MACHINE LEARNING BERBASIS DATA AKADEMIK (Studi Kasus : Universitas XYZ)

Abstract

Higher education is crucial for developing competitive human resources, yet the issue of student dropout (DO) remains a significant challenge for institutions. This study aims to develop a predictive model for identifying students at risk of dropout using machine learning techniques. By analyzing academic data, including Grade Point Averages (GPA), course loads, attendance rates, and failure rates, the research employs three machine learning algorithms: Decision Tree, Naive Bayes, and K-Nearest Neighbor (KNN). The results indicate that the Decision Tree model outperforms the others, achieving a perfect accuracy of 100% in classifying students as either "Graduated" or "Dropout." Naive Bayes also shows strong performance with 95% accuracy, particularly excelling in identifying actual dropout cases. Conversely, KNN exhibits the lowest effectiveness. The findings suggest that implementing the Decision Tree model can significantly enhance early detection and intervention strategies for at-risk students, ultimately improving academic management and student retention rates.