ISBN: 978-981-11-0008-6 DOI: 10.18178/wcse.2016.06.084
Neural Network Model Using Back-Propagation Algorithm with Momentum Term for Credit Risk Evaluation System
Abstract— This paper present the results of an experiment made with the aid of KangarooBPNN, a
graphical user interface software in order to find the Mean Squared Error (MSE) of a supervised neural
network model. In the previous experimentation or study, NN-1B model was considered to be a good neural
network model with 20 input neurons, 10 hidden neurons and 1 output neuron using 0.3 and 0.4 learning rate
and accuracy rate respectively at 10,000 epochs. The German credit dataset was used to train and test the said
neural network model using the back-propagation algorithm with momentum term for credit risk evaluation
system. The results were recorded in a tabular form, compared and analyzed carefully. Then it was compared
with the result of the previous study wherein the traditional back-propagation algorithm was used. Moreover,
based on the comparison made by the researchers, it shows that it is better to use the back-propagation
algorithm with momentum term than the traditional back-propagation algorithm.
Index Terms— neural network, neural network model, back-propagation algorithm, back-propagation algorithm with momentum, credit risk evaluation, credit scoring.
Dionicio D. Gante, Bartolome T. Tanguilig III
Technological Institute of the Philippines, PHILIPPINES
Bobby D. Gerardo
West Visayas State University, PHILIPPINES
Cite: Dionicio D. Gante, Bobby D. Gerardo, Bartolome T. Tanguilig III, "Neural Network Model Using Back-Propagation Algorithm with Momentum Term for Credit Risk Evaluation System," Proceedings of 2016 6th International Workshop on Computer Science and Engineering, pp. 500-506, Tokyo, 17-19 June, 2016.