PENERAPAN METODE NAÏVE BAYES DALAM MENGUKUR POTENSI KELULUSAN MAHASISWA BARU

Abstract

The high rate of delayed graduation remains a systemic challenge in Indonesian higher education, including at Universitas Medika Suherman (UMEDS). This research aims to build an Educational Data Mining (EDM)-based early warning system model to evaluate the readiness of new students for on-time graduation by implementing the Naïve Bayes Classifier algorithm.This study uses historical data of 184 students from the Faculty of Computer Science, UMEDS, from the 2016-2020 cohorts who have completed their studies. Seven predictor variables were analyzed: Grade Point Average of Semester 1 (GPA1), Attendance Rate, School Origin, Gender, Age, Payment Status, and Organizational Activity, with Graduation Status (On-Time vs. Delayed) as the target variable.The results identified GPA1 and Attendance Rate as the most significant predictive factors. Students with a GPA1 ≤ 2.75 have a 40% probability of graduating late, similarly to students with an attendance rate ≤ 80%. The constructed Naïve Bayes model showed highly optimal performance, with 94.6% accuracy on testing data, and precision, recall, and F1-Score of 97.2%. Validation using 10-Fold Cross Validation also confirmed the model's consistency and reliability with an average accuracy of 93.2%.