Developing a Predictive System for On-Time Graduation Using Logistic Regression

Abstrak

Timely graduation is widely recognized as a key indicator of academic quality and institutional effectiveness in higher education. While previous studies have examined individual predictors of student progression, few have combined academic, demographic, and socioeconomic factors into a comprehensive predictive model, particularly within the context of Indonesian private universities. This study aims to identify the main factors influencing on-time graduation by applying logistic regression to student data collected from a private university’s academic information system. The dataset includes 9,012 undergraduate records from cohorts entering between 2017 and 2020, covering a range of academic, admission, and background variables. The analysis reveals that fourth-semester GPA, attendance rate, scholarship status, completion of mandatory courses, and early course load have a significant impact on the probability of graduating on time. The predictive model achieved an accuracy of 85.76% and a recall of 90%, demonstrating strong classification performance. Although the findings are based on data from a single institution, the results offer practical insights for developing academic early warning systems and inform data-driven planning in higher education management.