Several proximity measures have been proposed to compare classifications derived from different clustering algorithms. There are few proposed solutions for the comparison of two classification trees; some of them measure the difference between the structures of the trees, some other compare the partitions associated to the trees taking into account their predictive power. Their features and limitations have been discussed; furthermore, a new dissimilarity measure has been proposed. It considers both the aspects explored separately by the previous ones. Three measures have been compared analyzing two different classification problems: a real data set and a simulation study. With respect to the real data set it has also been evaluated how and how much each of the considered measures is influenced by the presence of highly predictive variables which are also highly correlated.
The comparison between classification trees through proximity measures
MIGLIO, ROSSELLA;SOFFRITTI, GABRIELE
2004
Abstract
Several proximity measures have been proposed to compare classifications derived from different clustering algorithms. There are few proposed solutions for the comparison of two classification trees; some of them measure the difference between the structures of the trees, some other compare the partitions associated to the trees taking into account their predictive power. Their features and limitations have been discussed; furthermore, a new dissimilarity measure has been proposed. It considers both the aspects explored separately by the previous ones. Three measures have been compared analyzing two different classification problems: a real data set and a simulation study. With respect to the real data set it has also been evaluated how and how much each of the considered measures is influenced by the presence of highly predictive variables which are also highly correlated.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
