Bayesian Mixture Multinomial Regression Model for Loan Repayment Classification of Village Unit Cooperative

Abstract

Village Unit Cooperative (VUC) is an institute that provides financial services
consisting of several sub-businesses, one of which is providing loans. To downplay the
danger of non-performing loans, the cooperative must be capable to assess prospective
borrowers before a loan decision is required. In this case, loan repayment with its three
categories coupled with two types of the installment and interest payment systems is
constructed to create a suggested structure of Mixture Multinomial distribution with two
components. This construction is used to determine loan repayment of the applicant by
using the Bayesian Mixture Multinomial Regression model (BMMRM). The estimating
parameter is performed by making an algorithm based on Bayesian Markov Chain Monte
Carlo (MCMC) couple with the Gibbs sampling algorithm. The classification
performance of loan repayment through the BMMRM is assessed by determining loan
repayment classification percentage of the model which is compared with loan repayment
classification percentage of the Polytomous Logistic Regression Model (PLRM). The
comparative results show that the BMMRM approach gives a higher percentage of loan
repayment classification performance than the PLRM.