GIS Tool: Bayesian Networks

A Bayesian network - it’s complicated…

As discussed in Episode 12 of the podcast, a Bayesian network is a model that brings known factors as well as unknowns and degrees of certainty of your data into a prediction model. It’s pretty complicated (at least for me), but it may be useful to some of you who are more knowledgeable about probability theory!

A Bayesian network model is something mostly used in academia right now (see academic papers and journal articles here, here, and here, and even notes from a Stanford class here and an older tutorial here), but Esri, QGIS and others are incorporating tools to be able to more easily do this type of modeling inside of GIS software. 

We’ve got this!

A few Esri desktop software tools that use this model include Class Probability which is run against a raster image, Empirical Bayesian Kriging which requires a license of Geostatistical Analyst to run, and Maximum Likelihood Classification which requires the Spatial Analyst extension.

QGIS is a free, open source GIS software. There is a plugin called the PMAT (probabilistic map algebra tool) that users can download in order to integrate Bayesian network models with their geospatial data.

Mark Altaweel wrote a much better article over at GIS Lounge, so why not head over and read that for more information!

If I got anything majorly wrong or if you use this type of modeling in your work, please let us know in the comments below!