Credit risk evaluation using cycle reservoir neural networks with support vector machines readout

Abstract

Automated credit approval helps credit-granting institutions in reducing time and efforts in analyzing credit approval requests and to distinguish good customers from bad ones. Enhancing the automated process of credit approval by integrating it with a good business intelligence (BI) system puts financial institutions and banks in a better position compared to their competitors. In this paper, a novel hybrid approach based on neural network model called Cycle Reservoir with regular Jumps (CRJ) and Support Vector Machines (SVM) is proposed for classifying credit approval requests. In this approach, the readout learning of CRJ will be trained using SVM. Experiments results confirm that in comparison with other data mining techniques, CRJ with SVM readout gives superior classification results.