Evaluation of university accreditation prediction system

Abstract

Higher education accreditation body (Badan Akreditasi Nasional Perguruan Tinggi/BAN-PT in Indonesian) is an Indonesian body with the main task to assess the quality of Indonesian university. The assessment result is called accreditation, which has 5 years of validation time. In order to monitor the quality, University has an internal mechanism which is called an internal quality assurance system (Sistem Penjaminan Mutu Internal/SPMI in Indonesian). Usually, SPMI assesses the quality periodically, one or two times each year. This process needs much effort, i.e. time, manpower, and financial cost. Sometimes, internal auditor of the university does not have sufficient knowledge, as much as BAN-PT assessor. This condition causes a lack of assessment accuracy, then causes the quality of SPMI itself. On the other hand, University has abundant of condition data, saved in higher education database (HEDB). This paper proposes to exploit the availability of data in this case. Therefore university able to monitor the quality by machine learning process periodically without much effort as manual SPMI process. Furthermore, this paper evaluates two machine learning methods, i.e. naive Bayes and K-Nearest Neighbor (K-NN). This proposal exploits several data: student, academic, admission, and alumna. K-NN and naive Bayes work in registrant and capacity ratio, student registration ratio, average student Grade Point Average (GPA) in late five years, and on-time graduation scale. The experiment results show the average accuracy of naive Bayes and KNN are 70% and 95.2% respectively.