A Soft Computing Framework for Credit Risk Evaluation

Gafa Lubem, Agaji Iorshase, Esiefarienrhe. B. Micheal

Default Credit portfolios have historically been the major cause of financial distress in Nigeria commercial banks because of its inherent risk of possible loan losses (credit risk). It has also contributed to shareholders losing their investment in the commercial banks and inaccessibility of bank loans to the public. Also, evaluations of loan applications by Nigeria Banks are based on a loan officers' subjective assessment. Such assessment is inefficient, lack the ability to learn from customer past financial activities, inconsistent, bias in nature, delay in decision making, unreliable and non-uniform risk assessment which may lead to bankruptcies and defaulted loans. A soft computing framework based on a neurofuzzy model integrated with the customer Bank Verification Number (BVN) to link other Banks was developed. The method used deposit rate, withdrawal rate, average withdrawal, average deposit, loan request amount, collateral value, available cash and average income as inputs which was trained by the neural network to generate four output,that is, collateral, capital, character and capacity. Fuzzy rules were used to evaluate and determine the credit risk level for possible recommendation. The data for this framework was sourced from nine Nigeria Commercial banks. The soft computing framework was implemented using Java enterprise edition programming language. Experimental results showed that the system sufficiently evaluated the credit risk and generated the appropriate risk level with high accuracy for credit recommendation to minimized default loans in the Nigeria Banks.

https://doi.org/10.31871/IJNTR.5.12.32