Genetic algorithms for geographical boundary assignment

Presents a system for identifying near-optimal boundaries or regions on a map in order to maximise a set of boundary objectives. A genetic algorithm is used to perform the boundary allocation. A novel genetic encoding is described where sets of small spatial regions are represented in the genotype. The spatial areas are assembled from these sets of smaller spatial units during the decoding process. A new genetic operator, referred to as the minimal-mutation group transfer operator, is developed. This operator is designed to recombine grouping information in parents while introducing the least number of new foreign groupings when creating children. This results in a grouping crossover with a low effective mutation rate. This operator is found to be effective with the encoding scheme in producing a set of contiguous boundaries on a sample boundary problem