Applying niching genetic algorithms for multiple cluster discovery in spatial analysis

Traditional genetic algorithms with elitist selection are unable to locate more than one solution in a multimodal fitness landscape in a single run. This genetic drift is illustrated in Mapex, a smart spatial analysis technique, employing a genetic algorithm for spatial cluster discovery. However, for detecting multiple clusters Mapex provides a non-ideal approach. In this paper, we use a fitness sharing niching technique along with circle overlap detection to enhance Mapex. The results show the suitability of such an approach to locate solutions in a multimodal domain.