A statistical approach for the fine-tuning of metaheuristics: A case study combining design of experiments and racing algorithms

• Autor(a) principal: De Moraes Barbosa, Eduardo Batista [UNESP]
• Data de Publicação: 2015
• Outros Autores: Senne, Edson Luiz França [UNESP], Silva, Messias Borges [UNESP]
• Tipo de documento: Artigo de conferência
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://hdl.handle.net/11449/177959
• topic: Design of experiments; Fine-tuning; Metaheuristics; Racing algorithms; Response surface methodology

The fine-tuning of heuristics and metaheuristics exercises a great influence in both the solution process, as well as in the quality of results of optimization problems. The search for the best fit of these algorithms is a major research challenge in the field of metaheuristics. This paper aims to present a study on applying Design of Experiments (DOE) methodology combined with racing algorithms in the fine-tuning of different algorithms, such as Simulated Annealing (SA) and Genetic Algorithm (GA), to solve a classical scheduling problem. It will be presented the results comparison considering the default metaheuristics and ones using the settings suggested by the approach combining DOE and racing algorithm. Broadly, the proposed approach improves the quality of the solutions and allows for both GA and SA stay closer to optimum for different instances of the studied problem. Therefore, by means of this study it can be concluded that the combined use of DOE and racing algorithms may be a promising and powerful tool to assist in the investigation, as well as in the fine-tuning of different algorithms.