GIS software with Bayesian Network capabilities

Question

I am looking for any GIS software that has Bayesian Network capabilities. Ideally I'd be able to define grid layers as discrete or continuous inputs in to the network and then be able to produce layers based on the predictions of given nodes. I note there are several academic papers that seem to use a variety of different GIS and Bayesian tools but there doesn't seem to be any mention of an integrated solution. I'm starting to think that a solution doesn't exist and I'm going to hack my own solution together.

Answer 1

A tool created for ArcGIS has been done by the University of Queensland and Norsys : (this is a bit dated as was using ArcGIS 9.2)

This is a simple example for a model of land surface stability. The BN classification tool categorizes nodes as: background, classification and observation variables. In the figure above, the three BN variables for slope, soil and wetness index are influential factors (background variables) used to derive (classification variable) slope stability. The direct consequences of slope instability may be observed (observation variable) from surface failures such as soil creep or slips. The tool is designed to interpret classification problems following this generic structure, we believe many problems fit this form.

Norsys software was used:

http://www.norsys.com/netica.html

includes a free demo - http://www.norsys.com/netica.html#free

You can contact the University of Queensland for an update on the subject

http://www.gpem.uq.edu.au/cser-tools-bngis

http://www.gpem.uq.edu.au/contact-us