Design Principles and Performance Analysis of a Selection Hyper-heuristic across Multiple Problem Domains

OpenAlex · İngilizce

The present study is concerned with the design and analysis of a selection hyper-heuristic for solving different problems using different heuristic sets under different execution time limits. Several adaptive hyper-heuristic components were devised to provide high performance for the aforementioned cases. For properly combining these components, a number of coordination mechanisms were developed. The proposed approach composed of these elements was tested using a high-level search framework that currently consists of six of problem domains. The empirical results indicated that the method suggested is capable of adapting itself under different circumstances and efficiently coordinating its subcomponents. It delivered clear performance difference compared to a group of hyper-heuristics from the literature. In addition, it showed superior performance as the winner of an international competition on HyFlex, i.e. the Cross-domain Heuristic Search Challenge 2011.