Exploiting the Cost (In) Sensitivity of Decision Tree Splitting Criteria
- Chris Drummond, Institute for Information Technology, National Research Council Canada
- Robert Holte, Department of Computing Science, University of Alberta
This paper investigates how the splitting cri teria and pruning methods of decision tree learning algorithms are influenced by misclas sification costs or changes to the class distri bution. Splitting criteria that are relatively insensitive to costs (class distributions) are found to perform as well as or better than, in terms of expected misclassification cost, splitting criteria that are cost sensitive. Con sequently there are two opposite ways of deal ing with imbalance. One is to combine a cost insensitive splitting criterion with a cost in sensitive pruning method to produce a deci sion tree algorithm little affected by cost or prior class distribution. The other is to grow a costindependent tree which is then pruned in a costsensitive manner.
Citation
C. Drummond, R. Holte. "Exploiting the Cost (In) Sensitivity of Decision Tree Splitting Criteria". International Conference on Machine Learning (ICML), Stanford University, pp 239-246, January 2000.Keywords: | exploiting, tree, splitting, criteria, machine learning |
Category: | In Conference |
BibTeX
@incollection{Drummond+Holte:ICML00, author = {Chris Drummond and Robert Holte}, title = {Exploiting the Cost (In) Sensitivity of Decision Tree Splitting Criteria}, Pages = {239-246}, booktitle = {International Conference on Machine Learning (ICML)}, year = 2000, }Last Updated: June 18, 2007
Submitted by Staurt H. Johnson