Authors: T.J. O'Neill, Jack Penm
Addresses: School of Finance and Applied Statistics, The Australian National University, Canberra, ACT 0200, Australia. ' School of Finance and Applied Statistics, The Australian National University, Canberra, ACT 0200, Australia
Abstract: Conventional methods to test for credit ratings of financial debt issuers based on current means of classification are typically undertaken in the framework of applied statistical methods. In this paper, a newly introduced approach, Support Vector Machines (SVMs), has been applied to test a set of Standard & Poor (S&P)|s issuers| credit rating data. The primary purpose of this credit rating analysis is to measure the credit worthiness of credit securities| issuers and thus provide investors valuable information in making financial decisions. To construct our classification model, the ten key financial variables used by S&P|s, and a dummy country variable, are used as the input variables. A conventional full-order neural network based classification model is selected as the benchmark. Our findings indicate the superiority of the SVMs approach over the neural network approach.
Keywords: classification; credit ratings; financial services; financial standards; learning models; financial debt issuers; support vector machines; SVM; neural networks.
International Journal of Services and Standards, 2007 Vol.3 No.4, pp.390 - 401
Published online: 28 Sep 2007 *Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article