An application of a counter-propagation neural network: simulating the Standard and Poor's Corporate Bond Rating system

Abstract

Various neural network models have proven useful in vision and other sensory input pattern recognition applications. Much of the earlier work focused on military and defense. Neural network classification ability is just beginning to be deployed in financial applications. Some areas already explored with promising results are credit analysis, market analysis, fraud detection, and price forecasting. Elements in common between the military sensory input and the financial applications include huge volumes of data, time-critical processing, pattern complexity, and qualitative decision criteria. This paper covers research performed to build a Standard and Poor's corporate Bond Rating simulator using the unidirectional version of the counter-propagation network model invented by Robert Hecht-Nielsen (1988).<>