Mining Ecological Data with Cellular Automata

Abstract

This paper introduces a Cellular Automata (CA) approach to spatiotemporal data mining (STDM). The recently increasing interest in using Genetic Algorithms and other evolutionary techniques to identify CA model parameters has been mainly focused on performing artificial computational tasks such as density classification. This work investigates the potential to extend this research to spatial and spatiotemporal
data mining tasks and presents some preliminary experimental results.
The purpose is twofold: to motivate and explore an evolutionary CA approach to STDM, and to highlight the suitability of evolutionary CA models to problems that are ostensibly more difficult than, for example, density classification. The problem of predicting wading-bird nest site locations in ecological data is used throughout to illustrate the concepts,and provides the framework for experimental analysis.