Emulating the Prospector Expert System with a raster GIS

Abstract Recent enhancements to the MAPS public-domain raster geographic information system (GIS) allow it to emulate the methodology of the Prospector Expert System. This pragmatic implementation deviates from the classic methods of expert systems, forward and backward chaining, but uses the standard GIS method of cartographic modeling. This allows faster processing of cell maps, but requires that the rule base be ordered before use. There is some evidence in the original description of Prospector of deviation from the reported method of backward chaining. The Prospector methodology is discussed, and step-by-step examples of Bayesian inference, fuzzy logic, and certainty calculations are presented. Expert systems provide the ability to process a cartographic model with missing or incomplete data. A simplified model of the original Prospector rules determining a favorable location to drill for copper was used to test this methodology. This model can accept up to 26 maps as input. The example in this paper uses 23 maps. The results are compared with those of the original Prospector report.