G 9 . 6 The set covering problem

In this case study we consider the set covering problem, the classical operations research problem of covering the rows of a zero-one matrix by a subset of the columns at minimum cost. We outline a genetic-algorithm-based heuristic for the problem. The key features of this genetic algorithm are a binary representation, drawn naturally from a zero-one formulation of the problem, a new crossover operator (fusion), and a variable mutation schedule based upon the convergence of the algorithm without mutation and a heuristic operator to ensure feasibility (i.e. to ensure that each individual satisfies the problem constraints). We discuss computational results on a set of standard test problems drawn from the literature. These show that the genetic-algorithm-based heuristic is capable of finding better quality solutions than other approaches currently existing in the literature within a reasonable computation time.