On Fitness Landscape Analysis of Permutation Problems: From Distance Metrics to Mutation Operator Selection

In this paper, we explore the theory and expand upon practice of fitness landscape analysis for optimization problems over space permutations. Many computational analytical tools analysis, such as distance correlation, require identifying a metric measuring similarity different solutions to problem. We begin with survey available metrics permutations, then use principal component classify these metrics. The result aligns existing classifications permutation problem types produced through less formal means, including A-permutation, R-permutation, P-permutation types, which classifies by whether absolute position elements, relative positions or general precedence pairs is dominant influence solution fitness. Additionally, identifies subtypes within categories. see that classification can assist in appropriate based on feature analysis. Using each class, also demonstrate how scheme subsequently inform choice mutation operator an evolutionary algorithm. From this, present variety operators counterpart Our implementations metrics, operators, associated algorithm, are pair open source Java libraries. All code necessary recreate our experimental results source.