Driven PCTBagging: Seeking Greater Discriminating Capacity for the Same Level of Interpretability

The partial consolidated tree bagging (PCTBagging) was presented as a multiple classifier that, based on a parameter, the consolidation percentage, can exploit more the possibilities of the inner ensembles, and obtain higher levels of interpretability, or can exploit more the possibilities of the ensembles, and obtain higher discriminant capacity. Thus, at the extreme values, with a consolidation percentage of 100% it obtains a consolidated tree (CTC algorithm) and with 0% consolidation it obtains a Bagging. For intermediate values, the consolidated tree is collapsed to the number of internal nodes corresponding to the percentage value, selecting the biggest possible nodes. In this paper we propose a strategy to directly develop the partial consolidated tree, i.e. without the need to build the complete consolidated tree and, in addition, we explore up to 4 other different criteria, besides the size of the nodes, to decide which will be the next node to be developed in the partial consolidated tree: Pre-order, Gain ratio, Gain ratio \(\times \) Size, and, Level by level. The results show that the use of different criteria affects the discriminant capacity of the classifier for the same level of interpretability, and that this effect is greater the higher the percentage of consolidation is.