A grouping method for categorical attributes having very large number of values
MLDM'05 Proceedings of the 4th international conference on Machine Learning and Data Mining in Pattern Recognition
| Hi-index | 0.00 |
The tests associated with the internal nodes of a binary decision tree can take on a variety of forms; for instance, they can test whether a given attribute has a given value, or more generally, whether it has a value in a given set. This report examines whether allowing the latter form of test (which we call a value group) provides significant advantages over allowing only single-valued tests. The results of an empirical study comparing performance of these two approaches on a variety of learning tasks are reported. For the data sets studied, value grouping is shown to have no strong effect on classification accuracy, although it does result in smaller trees.