Naïve Bayes Classification for Predicting the Timeliness of Student Graduation: A Case Study at Universitas Sembilanbelas November Kolaka

Abstrak

Graduate on time is a key indicator of academic quality, yet many higher-education institutions in Indonesia continue to experience a persistent gap between student enrollment and on-time completion. At the Faculty of Information Technology of a Universitas Sembilanbelas November Kolaka in Southeast Sulawesi, only 17 of 188 Information Systems students admitted in 2015 graduated on schedule by 2019, indicating that a substantial proportion of students require remedial planning support. This study develops and evaluates a web-based classification system that applies the Naïve Bayes algorithm to predict whether a student is likely to graduate “On-Time” or “Delayed” based on seven categorical attributes: gender, grade point average for semesters I–IV, organizational involvement, and employment status. A dataset of 100 alumni records was collected, cleaned, and discretized, then split using the Hold-Out method into 70 training records and 30 testing records. The system was implemented in PHP with a MySQL backend following the Waterfall software development life cycle and validated against manual Bayesian probability calculations. Model performance was assessed using a confusion matrix, yielding an accuracy of 77%, a specificity of 91%, a sensitivity of 50%, and an error rate of 23%. Manual and system-based computations produced identical classification outcomes for all sampled cases, confirming correct implementation of the algorithm. Additional Hold-Out experiments using 60:40, 70:30, and 80:20 training-to-testing ratios showed that classification accuracy is sensitive to the amount of training data available. These findings demonstrate that Naïve Bayes, despite its strong feature-independence assumption, is a computationally efficient and sufficiently accurate method for early identification of students at risk of delayed graduation