Skip to content

Decision tree grafting

Grafting is a postprocess that adds tests and nodes to existing decision trees, resulting in bagging-like variance reduction while providing a single interpretable decision tree.  It was developed as part of my exploration of the use of Occam's Razor in machine learning.

The first grafting algorithm added new branches that did not change the tree's classification of any training data [Webb, 1996].  This constraint was imposed so as to create as clear a demonstration that it was not true that all other things being equal a less complex classifier should be more accurate than a more complex one.  Subsequent algorithms [Webb, 1998, 1999] were intended as practical approaches to machine learning and allowed branches that reclassified training examples that the original tree misclassified.

The original decision tree grafting software was implemented as an extension to C4.5.  It can be downloaded here (requires C4.5 release 6).  Decision tree grafting is also implemented in the J48Graft component of the Weka machine learning workbench.

Decision tree grafting was first conceived as a way of demonstrating that of two models that perform identically on training data, the simplest is not most likely to have the lowest accuracy.  However, it has also proven to be a practical learning algorithm. Indeed, its Weka implementation has been used in a number of scientific applications, including those listed here.

Publications

Webb, G. I.

Decision Tree Grafting From The All Tests But One Partition

Dean, T. (Ed.): Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence (IJCAI 99), pp. 702-707, Morgan Kaufmann, Stockholm, Sweden, 1999.

Abstract | BibTeX

Webb, G. I.

The Problem of Missing Values in Decision Tree Grafting

Antoniou, G.; Slaney, J. K. (Ed.): Lecture Notes in Computer Science Vol. 1502: Advanced Topics in Artificial Intelligence, Selected Papers from the Eleventh Australian Joint Conference on Artificial Intelligence (AI '98), pp. 273-283, Springer-Verlag, Brisbane, Australia, 1998.

Abstract | BibTeX

Webb, G. I.

Decision Tree Grafting

Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence (IJCAI 97), pp. 846-851, Morgan Kaufmann, Nagoya, Japan, 1997.

Abstract | BibTeX

Webb, G. I.

Further Experimental Evidence Against The Utility Of Occam's Razor

Journal of Artificial Intelligence Research, vol. 4, pp. 397-417, 1996.

Abstract | Links | BibTeX