Radial Basis Function Neural Networks in Credit Application Vetting Systems
This paper describes an investigation carried out at Coventry into the suitability of Radial Basis Function(RBF) Neural Networks for use in credit vetting systems. The RBF used is an All Classes in One configuration with unweighted Euclidean distance measure. The paper examines the performance of the RBF network in classifying good and bad loan cases over a data set supplied by a substantial finance organisation. The effects of changing the number of centres and the training regime are examined. The network prediction performance is compared over the same data set with both a manually configured, and a genetic algorithm designed Back Propagation Trained Multi-Layer-Perceptron.
The performance of the RBF network in classifying cases is found to be comparable with these two solutions, providing similar prediction rates. The relative simplicity of the RBF solution gives greatly reduced computing time for comparable performance, and potentially easier routes to providing information about the decisions reached. Ideas for enhancing the future performance of the system are discussed.
KeywordsRadial Basis Function Singular Value Decomposition Radial Basis Function Neural Network Radial Basis Function Network Cerebellar Model Articulation Controller
Unable to display preview. Download preview PDF.
- 1.Refenes A.N. Neural Network Applications in Investment and Finance Services. Chapter 27, E. Turban, R Trippi. USA. Probus Publishing. 1992.Google Scholar
- 2.Refenes A.N., Managing Exchange Rate Prediction Strategies with Neural Networks. In Techniques and Applications of Neural Networks, Chapter 7, Taylor and Lisboa (eds). Chichester. Ellis Horwood. 1993.Google Scholar
- 4.Roy J. Suret J.M. A clever screening system for commercial loan applications. In Proceedings of the Conference on Expert Systems in Economics, Banking and Management. 1989.Google Scholar
- 5.Bazley G. Report of the Practical Application of a Neural Network in Financial Service Decision Making. In Proceedings of the International Conference on Artificial Neural Nets and Genetic Algorithms. Innsbruck, Vienna. Springer Verlag. 1993.Google Scholar
- 6.Williamson A.G. Fitness Criteria for the Genetic Algorithm Optimization of a Neural Network Credit Application Vetting System In Proceedings of the Conference on Financial Information Systems. Sheffield Hallam University. Sheffield. 1994.Google Scholar
- 7.Wettschereck D. Dietterich T. Improving the performance of Radial Basis Function Neural Networks by Learning Centre Locations. In Proceedings of the Conference on Neural Information Processing Systems-Natural & Synthetic. 1991.Google Scholar
- 9.Wong Y. How Gaussian Radial Basis Functions Work. In Proceedings of the International Joint Conference on Neural Networks. Seattle. 1991.Google Scholar
- 10.Carpenter G.A Grossberg S. Self-organization of stable category recognition code for analog input patterns. In Proceedings of the IEEE International Conference on Neural Networks. San Diego. IEEE. 1987.Google Scholar
- 11.Musavi M.T. Faris K.B. Chan K.H. Ahmed W. On the Implementation of Neural Network Applications. In Proceedings of the Conference on Analysis of Neural Networks Applications. 1991Google Scholar