Genetic Algorithms for a Large Scale Dynamic Allocation Problem

Mate-selection is the problem of deciding which animal should be culled and which should be mated in a breeding program. In addition, if the animal is to be mated, which animal from the other sex should it be mated with? With a linear objective function, integer linear programming was successfully used to solve the problem. When the relation between the animals' traits is additive, a strategy based on the classical index theory for selection with random allocation results in the optimal solution. Today, non-linear and non-additive objectives are introduced and the problem is becoming increasingly complex. In this paper, we formulate the problem as a multi-stage quadratic transportation model. We then solve the problem using two versions of Genetic Algorithms (GAs) on data collected from the Australian Dairy Industry. We then compare GAs against two conventional heurstic techniques; these are random search and simulated annealing. GAs are found to be better than the other heruistics for handling this large scale mate-selection problem