How can I use a neural network in GIS?

Question

I'd like to learn how to use a neural network in GIS. I surfed in the internet to find good resources or a tutorial about this but I couldn't find any helpful resources. Are there any good references or tutorials about using neural networks in GIS? I'd like to use an artificial neural network (ANN) in mapping hazard zones, for example flood zones or landslides. I use ArcGIS 10 to process the layers but don't know how can use an ANN.

Answer 1

Lee and Evangelista (2006) have a good article on earthquake-induced landslide-susceptibility mapping using an artificial neural network. They appear to do their GIS analysis in ArcGIS and implement a ANN algorithm in Matlab.

Spatial Data Modeler (SDM) is available as a collection of geoprocessing tools for ArcGIS (earlier Arc 9.x version here). The author describes SDM NN tools as follows:

The tools include the data-driven methods of Weights of Evidence, Logistic Regression, and two supervised and one unsupervised neural network methods, and a knowledge-driven method Fuzzy Logic.

If you find NNs useful for your projects, I suspect you will want to utilize the functionality of NNs in R. The following are links to useful sites and packages:

Overview [technical] article in the R Journal

Package ‘nnet’

Package ‘AMORE’

Package ‘neuralnet’