In this paper we present a genetic algorithm-based heuristic for non-u
nicost set covering problems. We propose several modifications to the
basic genetic procedures including a new fitness-based crossover opera
tor (fusion), a variable mutation rate and a heuristic feasibility ope
rator tailored specifically for the set covering problem. The performa
nce of our algorithm was evaluated on a large set of randomly generate
d problems. Computational results showed that the genetic algorithm-ba
sed heuristic is capable of producing high-quality solutions.