Abstract
In this paper, we discuss solution methods for a real world manpower allocation problem. The problem includes more than 7, 000 0-1 variables rendering the branch and bound method infeasible when considering the huge amount of computation time. We consider an application of the genetic annealing(GAn)which is an extension of genetic algorithm(GA)and simulated annealing(SA). In order to confirm that an optimal solution or a sub-optimal solution can be obtained GAn, using small-sized problems, the GAn solutions are compared with the solution of the branch and bound method. At the same time, we compare the obtained solutions with GA's and SA's. As a result, it is shown that the GAn solutions are optimal or at least sub-optimal. Appying the algorithms to the real size problem, it is shown that GAn produces a better solution. To obtain a further better solution, a diversification technique is introduced and its advantage is confirmed. Moreover, applying GA with GAn's genetic operators and GAn with GA's genetic operators to the problem, we analyzed the reason behind the success of GAn where GA failed.
