Dynamic aspects in neural classification

Author: Lanquillon C.^{1, *}

Source: International Journal of Intelligent Systems in Accounting, Finance & Management, Volume 8, Number 4, December 1999 , pp. 281-296(16)

Publisher: John Wiley & Sons, Ltd.

Buy & download fulltext article:

The full text article is not available for purchase.

The publisher only permits individual articles to be downloaded by subscribers.

Abstract:

When using neural networks for classification in an environment that is changing as time proceeds, methods for updating the parameters of the neural network should be considered in order to retain classification accuracy. Otherwise the neural network may lose performance due to structural changes. A classifier could be completely relearned from scratch at regular intervals. However, our experience from past credit scoring applications shows that users commonly prefer systems that change in as few cases as possible. Furthermore, this approach may be wasteful regarding the required computing time and that previously learned information will always be discarded. Therefore, we favor a methodology that attempts to detect changes and adapts a classifier only if inevitable. In this article, some methods for detecting and treating structural changes are applied to a credit scoring application. The results show that these methods may successfully be applied in a dynamic setting. Copyright © 1999 John Wiley & Sons, Ltd.

Language: English

Document Type: Research article

DOI: http://dx.doi.org/10.1002/(SICI)1099-1174(199912)8:4<281::AID-ISAF175>3.0.CO;2-7

Affiliations: 1: DaimlerChrysler Research and Technology, Germany *

Publication date: 1999-12-01

More about this publication?

As of January 2010, this journal will no longer be available on IngentaConnect, please visit Wiley InterScience to arrange continued access to this content
Access this journal at Wiley InterScience
ingentaconnect is not responsible for the content or availability of external websites