Pruning Multivariate Decision Trees by Hyperplane Merging
Abstract
Several techniques for induction of multivariate decision trees have been published in the last couple of years. Internal nodes of such trees typically contain binary tests questioning to what side of a hyperplane the example lies. Most of these algorithms use cut-off pruning mechanisms similar to those of traditional decision trees. Nearly unexplored remains the large domain of substitutional pruning methods, where a new decision test (derived from previous decision tests) replaces a subtree. This paper presents an approach to multivariate-tree pruning based on merging the decision hyperplanes, and demonstrates its performance on artificial and benchmark data.
Cite
Text
Kubat and Flotzinger. "Pruning Multivariate Decision Trees by Hyperplane Merging." European Conference on Machine Learning, 1995. doi:10.1007/3-540-59286-5_58Markdown
[Kubat and Flotzinger. "Pruning Multivariate Decision Trees by Hyperplane Merging." European Conference on Machine Learning, 1995.](https://mlanthology.org/ecmlpkdd/1995/kubat1995ecml-pruning/) doi:10.1007/3-540-59286-5_58BibTeX
@inproceedings{kubat1995ecml-pruning,
title = {{Pruning Multivariate Decision Trees by Hyperplane Merging}},
author = {Kubat, Miroslav and Flotzinger, Doris},
booktitle = {European Conference on Machine Learning},
year = {1995},
pages = {190-199},
doi = {10.1007/3-540-59286-5_58},
url = {https://mlanthology.org/ecmlpkdd/1995/kubat1995ecml-pruning/}
}