Download this Paper Open PDF in Browser

Classification Trees for Problems with Monotonicity Constraints

39 Pages Posted: 18 Feb 2003  

Rob Potharst

Erasmus University Rotterdam - Department of Computer Science; Erasmus Research Institute of Management (ERIM)

A. J. Feelders

University of Utrecht

Date Written: 23 2002 4,

Abstract

For classification problems with ordinal attributes very often theclass attribute should increase with each or some of theexplaining attributes. These are called classification problemswith monotonicity constraints. Classical decision tree algorithmssuch as CART or C4.5 generally do not produce monotone trees, evenif the dataset is completely monotone. This paper surveys themethods that have so far been proposed for generating decisiontrees that satisfy monotonicity constraints. A distinction is madebetween methods that work only for monotone datasets and methodsthat work for monotone and non-monotone datasets alike.

Keywords: monotone, monotonicity constraint, classification tree, decision tree, ordinal data

JEL Classification: M, M11, R4, C6

Suggested Citation

Potharst, Rob and Feelders, A. J., Classification Trees for Problems with Monotonicity Constraints (23 2002 4,). ERIM Report Series Reference No. ERS-2002-45-LIS. Available at SSRN: https://ssrn.com/abstract=370989

Rob Potharst (Contact Author)

Erasmus University Rotterdam - Department of Computer Science ( email )

P.O. Box 1738
Room T08-21
3000 DR Rotterdam, 3000 DR
Netherlands
+31 10 4081343 (Phone)
+31 10 4089167 (Fax)

Erasmus Research Institute of Management (ERIM)

P.O. Box 1738
3000 DR Rotterdam
Netherlands

A. J. Feelders

University of Utrecht ( email )

P.O. Box 80089
NL-3508 TC Utrecht
Netherlands

Paper statistics

Downloads
73
Rank
273,310
Abstract Views
823