Abstract

The objective of this research is to design an efficient mechanism to evaluate fault in a student loan approval process using the clustering technique with a concrete number of clusters. In this study, K-mean is used to analyze the historical data sets of student loan candidates provided by an anonymous university, collected between 2014 and 2015. There are 1,146 samples within two years of application. The result shows that, when k = 2, K-mean performs its best partitioning by categorizing the results by validating three well-known clustering measurement indexes. It also performs the best when comparing the overall average distance, at 4.41, with the highest density around the centroid of the cluster. This proposed model can identify errors in using the traditional selection criteria for granting student loans, which could lead to redesigning student loan approval criteria together with the development of a mobile application to support the administrators and executives who make student loan decisions.