CSNL: A cost-sensitive non-linear decision tree algorithm

Vadera, S ORCID: https://orcid.org/0000-0001-6041-2646 2010, 'CSNL: A cost-sensitive non-linear decision tree algorithm' , ACM Transactions on Knowledge Discovery from Data (TKDD), 4 (2) , pp. 1-25.

Preview

PDF - Accepted Version Download (770kB) | Preview

Abstract

This article presents a new decision tree learning algorithm called CSNL that induces Cost-Sensitive Non-Linear decision trees. The algorithm is based on the hypothesis that nonlinear decision nodes provide a better basis than axis-parallel decision nodes and utilizes discriminant analysis to construct nonlinear decision trees that take account of costs of misclassification. The performance of the algorithm is evaluated by applying it to seventeen datasets and the results are compared with those obtained by two well known cost-sensitive algorithms, ICET and MetaCost, which generate multiple trees to obtain some of the best results to date. The results show that CSNL performs at least as well, if not better than these algorithms, in more than twelve of the datasets and is considerably faster. The use of bagging with CSNL further enhances its performance showing the significant benefits of using nonlinear decision nodes. The performance of the algorithm is evaluated by applying it to seventeen data sets and the results are compared with those obtained by two well known cost-sensitive algorithms, ICET and MetaCost, which generate multiple trees to obtain some of the best results to date. The results show that CSNL performs at least as well, if not better than these algorithms, in more than twelve of the data sets and is considerably faster. The use of bagging with CSNL further enhances its performance showing the significant benefits of using non-linear decision nodes.

Item Type:	Article
Uncontrolled Keywords:	decision tree learning, cost-sensitive learning, machine learning, AI
Themes:	Subjects / Themes > Q Science > QA Mathematics > QA075 Electronic computers. Computer science Subjects outside of the University Themes
Schools:	Schools > School of Computing, Science and Engineering Schools > School of Computing, Science and Engineering > Salford Innovation Research Centre
Journal or Publication Title:	ACM Transactions on Knowledge Discovery from Data (TKDD)
Publisher:	ACM Digital Library
Refereed:	Yes
ISSN:	1556-4681
Depositing User:	S Vadera
Date Deposited:	24 Jun 2010 08:32
Last Modified:	21 Dec 2018 17:10
URI:	http://usir.salford.ac.uk/id/eprint/9406

Actions (login required)

Edit record (repository staff only)