Exploiting Fitness Distance Correlation of Set Covering Problems
The set covering problem is an NP-hard combinatorial optimization problem that arises in applications ranging from crew scheduling in airlines to driver scheduling in public mass transport. In this paper we analyze search space characteristics of a widely used set of benchmark instances through an analysis of the fitness-distance correlation. This analysis shows that there exist several classes of set covering instances that show a largely different behavior. For instances with high fitness distance correlation, we propose new ways of generating core problems and analyze the performance of algorithms exploiting these core problems.
Unable to display preview. Download preview PDF.