Estimation of Loan Repayment Events in Microfinance Bank Using L1 - Lasso Penalized Cox Proportional Hazards Approach

Abstract

This study applied L1- Lasso estimation for Cox proportional hazards model to select variables that are relevant to credit repayment rates of loan by bank customers and to build a predictive model. A dataset was used to predict credit repayment time by the customers using L1- Lasso and Cox proportional hazards procedures in order to find the variables that are related to time to credit repayment rates, for building a sparse model and for predicting the survival of credit repayment rates by customer in the future. Records of 186 customers of a Microfinance bank is used. The L1 - Lasso penalized Cox proportional hazards approach was able to identify the most predictive variables for repayment rates of loan. The findings of this study have shown that, the repayment rates of loan is significantly related with marital status, type of collateral and residence, whereas the variables age and occupation are less significantly related with the survival of loan repayment rates of customers. Finally, the variables selected by the model can be used in granting loan to customers in Microfinance banking. R Programming Language was used for the analysis.