Non-linear principal component analysis-based hybrid classifiers: an application to bankruptcy prediction in banks
by Vadlamani Ravi, Chelimala Pramodh
International Journal of Information and Decision Sciences (IJIDS), Vol. 2, No. 1, 2010

Abstract: This paper presents various non-linear principal component analysis (NLPCA)-based two-phase hybrid classifiers for predicting bankruptcy in banks. The first phase of the hybrids performs dimensionality reduction using NLPCA, which is implemented as a threshold accepting trained auto associative neural network (TAAANN). By considering the non-linear principal components as new inputs, second phase is invoked. In the second phase, which is essentially a classifier, we employed threshold accepting neural network (TANN), TANN without hidden layer, threshold accepting trained logistic regression (TALR) and multi layer perceptron (MLP). The results are compared with that of MLP, radial basis function neural network and found that the proposed hybrids performed well. It was observed that the NLPCA-TANN hybrid outperformed other hybrids over all data sets studied here. Further, TALR outperformed all the hybrids over all data sets. Based on the results, we infer that the hybrid classifiers performed very well by yielding high accuracies.

Online publication date: Wed, 02-Dec-2009

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Information and Decision Sciences (IJIDS):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com