Using Genetic Learning Neural Networks for Spatial Decision Making in GIs
Jinjin Zhou, Daniel L. Civco · Photogrammetric Engineering & Remote Sensing · 1996
the traditional gradient descent-based backpropagation. 1 Traditional approaches for suitability analysis in GIS are Through evolutionary learning from samples, a neural netoverlay ad the more complicated multicriteria evaluation work adapts its connection weights to approximate the de(m~). Despite being widely used, these methods have at sired output. Then, a successfully trained neural network can least three problems: (11 difficulties in handling spatial ,jato accomplish the suitability analysis task. A set of experiments possessing inaccumcy, multiple measurement scales, and is presented, The results show that the difficulties in tradifactor interdependency; (2) requirements of prior knowledge tional are by evO1utionq learning and the in identihng criteria, assigning scores, determining criteria abfiity the neural network. preference, and selecting aggregation functions; and (3) typically, an unfriendly user inte$ace. TO solve these prob- lntroduction to Traditional Methods lems, in this paper a neural network approach is presented. Overlay The neural network uses a genetic algorithm as its learning The application of digital map overlay for the purpose of mechanism. A set of experiments revealed that the afore- identifying suitable areas is a classic application of GIs. In mentioned dijlficulties are overcome by the evolutionary raster GIS, for example in IDRISI (Eastman, 1995), a suitability learning of neural networks. Our conclusion is that genetic map is produced from a series of Boolean images, where learning neural networks can provide an alternative for and each image represents all areas meeting the criterion being improvement over traditional suitability analysis methods in depicted. These images are then combined the overlay GIS. combination procedure to yield a final map that shows the sites meeting all the specified criteria. However, overlays