Comparison of use of linkage in integrated cluster with discriminal analysis approach

Credit is a provision of money or bills based on a loan agreement between the bank and another party that requires the borrower to pay off the debt after a certain period with interest. Before the bank gives credit to the debtor, it needs an assessment from the bank to measure whether the debtor can fulfill his obligations in a credit or not. One of the credit problems is the existence of debtors who have non-current credit so that it can harm the bank. Of these problems, of course, there needs to be supervision in terms of credit, one of the statistical analyses that can be used on these problems is discriminant analysis. Discriminant analysis has several shortcomings. One problem that arises in discriminant analysis is if the sample obtained, has sub-samples that each member in the sub-sample has the same characteristics. If this happens the accuracy of the model in the discriminant analysis will be low. How to integrate cluster analysis with discriminant analysis is to use dummy variables obtained from the cluster results. In this study examines the application of integrated clusters in discriminant analysis with three different linkage methods, namely single linkage, complete linkage and average linkage. This study also wants to get the best linkage for use in integrated cluster discriminant analysis approaches that maximize hit ratio, sensitivity and specivisity. The hit ratio, sensitivity and specificity of the integrated cluster with average linkage, complete linkage and single linkage are better than discriminant analysis.