Autonomous rule generation and assessment for complex spatial modeling

Traditional geographic information systems (GISs) have provided a mechanismto allocate the individual weightings of physical variables for single spatial modeling applications. However, as the complexity of the decision-making process increases, as in conflicting resource allocation modeling, examination of many more complex resource and non-resource relationships becomes tedious and extremely difficult. What is needed is a technique to assist the modeler in effectively, and correctly, defining and evaluating large rule sets for complex spatial analysis. In this paper we will briefly review traditional techniques for defining rules used in combining large variable sets, outline a approach using genetic algorithms and neural network techniques for automated rule generation, and discuss the implementation of this approach in Hoosier National Forest. Results of this study clearly demonstrate the potential use of this approach for modeling complex resource problems.