Understanding evolutionary induction of decision trees
Krzysztof Jurczuk, Marcin Czajkowski, Marek Krętowski · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2021
This paper concerns the evolutionary induction of decision trees (DT)s. Noting that many DTs or their parts reappear during the evolution, a multi-tree concept is introduced. A multi-tree integrates all trees with the same test in the root. The differences between trees (below the root) are saved as variants, while the same DT parts are shared. All multi-trees are stored in an external repository that keeps the entire search history. We show that such a repository can achieve not only the goal of duplicated calculation prevention, but also better understanding of the evolutionary DT induction. The road to finding the best individual is easily retrieved, exploitation vs. exploration can be addressed, or similarities between individuals can be measured.