ABSTRACT
Many applications in geographical analysis require similar items to be grouped together according to some measure of their type and location. Once clustered, these items can be referred to by a label that is defined to embody the concept used to differentiate between the types. Genetic Algorithms (GAs) provide an approach that has no pre-determined bias towards particular sizes or shapes of clusters. They have been used to solve a wide range of problem types, including the induction of classifications from examples, and are even beginning to appear in commercial applications. In the spatial GA information from the problem domain is used to guide the generation of new solutions as well as providing the data to evaluate them. The adaptation of the GA paradigm reduces the amount of computation required for the class of spatial clustering problems by tailoring a weak, generally applicable optimisation method with problem specific heuristics.
