Hybrid neural systems for large scale credit risk assessment applications

Hybrid Neural Systems that integrate symbolic algorithms or fuzzysystems to Artificial Neural Networks (ANN) are a potential alternativeto the more traditional ANN models. However, in contrast with the ANNmodels, these systems have not been yet fully explored from a practicalviewpoint to show their effectiveness in large scale applications. Thispaper presents an extensive comparative analysis of the neuro-fuzzymodels FWD (Feature-Weighted Detector) and FuNN (Fuzzy Neural Network),together with their rule extraction techniques in a large-scale problem.Two aspects are considered: generalization performance of the models,and the interpretation and explanation qualities of the extractedknowledge. The experiments are conducted in the context of a large scalecredit risk assessment application in a real-world operation of aBrazilian financial institution. The results attained are compared tothose observed with multi-layer perceptron networks.