Output details

Generalised bottom-up pruning: A model level combination of decision trees

<24> This research generalises the bottom-up tree pruning approach to a multiple tree context, allowing to simultaneously simplify and combine the structure of multiple decision trees. Suitable pruning criteria were introduced and shown to produce significantly more compact (and similarly or better-performing) trees on a number of publicly available datasets, compared to single-tree approaches. The compactness of a tree is important particularly in an industrial context, as smaller trees are more easily interpreted and acted upon.