Credit risk prediction using neural network backpropagation algorithm

Abstract

Credit is a business activity that contains high risk and greatly affects the health and survival of banking businesses and financing institutions. Predicting credit risk is very beneficial for banks or financing institutions in taking decisions to establish credit. Decision makers of banks and financing institutions must have the precautionary principle to minimize credit risk when credit will be provided. The study designed credit risk prediction software with artificial neural network methods backpropagationalgorithms. Artificial neural network backpropagation with 1 hidden layer and the amount of data for training and testing as many as 20 pieces consisting of 5 models and using the logsig activation function is able to predict credit risk with a truth percentage of 70%-80%. Training and testing is used using matlab 6.1 software. Based on these results, the study recommends the development of artificial neural network algorithms as an effective method on credit risk prediction systems.