Simplifying classification trees through consensus methods

Methods for comparing and combining classification trees based on proximity measures have been proposed in the last few years. These methods could be used to analyse a set of trees obtained from independent samples or from resampling methods like bootstrap or cross validation applied to the training sample. In this paper we propose, as an alternative to the pruning techniques, a consensus algorithm that combines trees obtained by bootstrap samples. The consensus algorithm we consider is based on a dissimilarity measure recently proposed. Experimental results are provided to illustrate, in two real data sets, the performances of the proposed consensus method.