Development of a Credit Default Model using R and Neural Network

Abstract

Lenders decide for the approval or rejection of debt proposals by following a credit rating procedure. The complex nature of existing rating process leads to unpleasant decisions by lenders and borrowers. Therefore, lenders are struggling to find flexible and simple rating methods which are widely acceptable, comprehensive, objective and modifiable as per the lender's requirement [1]. Credit rating reflects the creditworthiness of borrowers [2]. A model based on financial data can provide more objectivity and flexibility to determine such creditworthiness. We have developed a model based on financial data of 100 companies from India. This model is developed in R and Neural network. This model can be used to predict whether the company will default in future or not. By training the model on 70% of the data we obtained an accuracy of 70.58%. Testing the model using remaining 30% of the data generates an accuracy of 68.75. The use of advanced techniques such as R and Neural networks coupled with financial data, makes this model comprehensive. Furthermore, this model saves time and sources while ensuring the accuracy of prediction. The proposed model could help to build, a reasonable system that can predict creditworthiness. This study provides a feasible future research scope.