Título: Graph -based comparison and analysis of metaheuristics
Autores: Maria Clara Ruas, Mateus Bastos Soares, João Pedro A.F. Campos & Michel Bessani
Resumo: The comparison of metaheuristics is a challenge in the optimization community. While searching for a solution, metaheuristics can have different forms of exploring the search space and mapping the objective space to find the best solution. Often, the behaviour of these methods can be quite unclear for direct human interpretation, making the comparison and selection of available methods subjective. Using graph theory and graph-related metrics, we can improve the understanding of how different metaheuristics estimate the optimal solution and have clear metrics for performance comparison. This leads to a more efficient and explainable approach for choosing the most suitable method for each problem and scenario. In this paper, we present the application of a graph-based representation and analysis of metaheuristics’ behaviour and discuss the interpretation of complex network metrics applied to compare the different metaheuristic methods, showing aspects relevant to the choice of which method is better for a specific problem. A case study is presented to illustrate the comparison between two metaheuristics. Results highlight how the graph-based comparison explains how the algorithms map the objective space in a visual and trackable manner, while the complex network metrics allow a direct comparison between the behaviour of the two different metaheuristics.
Palavras-chave: metaheuristics; optimization; graph; search trajectory network; complex networks.
Páginas: 8
Código DOI: 10.21528/CBIC2025-1173692
Artigo em PDF: CBIC_2025_paper1173692.pdf
Arquivo BibTeX:
CBIC_2025_1173692.bib