Output details
11 - Computer Science and Informatics
University of Kent
A hyper-heuristic evolutionary algorithm for automatically designing decision-tree algorithms
<24> This paper proposes the first evolutionary algorithm for automatically designing a full decision tree induction algorithm, inaugurating a new research topic at the intersection of evolutionary algorithms and decision tree induction. The automatically-designed decision tree algorithm has outperformed (with statistical significance) two very popular manually designed decision-tree algorithms (C4.5 and CART) across 20 classification datasets. This paper received the best paper award of three tracks of the ACM GECCO-2012 conference (flagship conference on evolutionary algorithms): Integrative Genetic and Evolutionary Computation, Self-* Search and Search-based Software Engineering, and an invited extended version is in press in the Evolutionary Computation journal.