Credit scoring using the hybrid neural discriminant technique

摘要: Abstract Credit scoring has become a very important task as the credit industry been experiencing double-digit growth rate during past few decades. The artificial neural network is becoming popular alternative in models due to its associated memory characteristic and generalization capability. However, decision of network's topology, importance potential input variables long training process often criticized hence limited application handling problems. objective proposed study explore performance by integrating backpropagation networks with traditional discriminant analysis approach. To demonstrate inclusion result from would simplify structure improve accuracy designed model, tasks are performed on one bank card data set. As results reveal, hybrid approach converges much faster than conventional model. Moreover, accuracies increase terms methodology outperform logistic regression approaches.