Locating optimal sites on raster suitability maps is a complex problem when the size of the sites is larger than the cell size. There are a number of techniques for generating suitability maps, but these maps are not solutions to the site location problem. Feasible solutions are clusters of contiguous cells which meet spatial criteria concerning the size and shape of the clusters. Locaring optimal sites involves a trade-off between the intrinsic suitability of individual cells and the spatial configuration of cells. This paper describes a genetic algorithm which searches for optimal clusters and thereby locates optimal sites. The genetic algorithm uses a parameterized regiongrowing program to translate numeric strings into regions on a raster and employs raster GIS functions to evaluate utility scores. In a range of simulated problems, comparisons with an exhaustive search show that the genetic algorithm is efficient and robust. The results indicate that the genetic algorithm can find good solutions to real problems when exhaustive search methods are impractical.
|Number of pages||12|
|Journal||Transactions in GIS|
|Publication status||Published - 1997|