Prevention of Non-Performing Assets in Banks : An Application of Artificial Neural Network

Over the last few years' Indian banking, in its attempt to integrate itself with the global banking, have been trying to reduce the Non-performing assets (NPAs) and to a great extent been successful in doing so. However, designing an intelligent Decision Support System for NPAs management can act as a major catalyst in efforts to reduce their incidence at the initial stage. Based on a primary survey of 100 bank managers engaged in sanctioning of industrial loan to small industries the Knowledge Based Decision Support System (KBDSS) is devised with the help of artificial neural network. Initially appropriate criteria (financial ratios and qualitative criteria) are identified and selected that are the most appropriate for the evaluation of the loan applications from borrowers. Each of the 26 selected criteria is modeled using four-point scale. Using Matlab Neural network toolbox, Perceptron and probabilistic neural networks are designed and trained with the set of training data. The network developed is able to discern the quality of the borrower and eventually draws inference such as "Loan can be granted", "Loan can not be granted" and "Doubtful cases". The developed network is tested with a set of 100 borrowers' data for its validation and further refinement. It is observed that the Probabilistic neural network provides a better decision in comparison to Perceptron network in identifying the three categories of loan applications. This approach to handle loan applications is a prototype and can be extended further to other loan categories which could be a preventive measure to the NPAs problem in Indian banking