Using Genetic Algorithms in Clustering Problems
Marco Painho and Fernando Ba??o
In this paper a Genetic Algorithm (GA) approach to the clustering problem is proposed. The optimisation capabilities of Genetic Algorithms are well known and commonly used in a variety of scientific fields. Openshaw and Openshaw (1997) note that Genetic Algorithms are "an extremely powerful, widely applicable search technique that provides a global search for problems with many local suboptima?". Here the clustering task is tackled through a Genetic Algorithm, which attempts to minimize the within cluster variance.
The Genetic Algorithm is tested against the traditional K-Means method, and an unsupervised neural network (Kohonen's self organising map). First, through the use of artificial generated data a particular environment is set up, then a real world example, involving the electoral results in Portugal, is used. Another important aspect is related with the possibility of using GA in geographical clustering problems. It seems obvious that the flexibility offered by GA should encourage researchers to develop new and improve ways of tackle the clustering tasks in Geography
Reference: Proceedings of the Fifth International Conference on GeoComputation
, University of Greenwich's School of Earth and Environmental Sciences, Kent, UK, 23 - 25 August 2000. Papers published on CD-ROM. Produced by: R.J.Abrahart and B.H.Carlisle. Publisher: "GeoComputation CD-ROM". ISBN 0-9533477-2-9