Application of the PROSPECTOR system to geological exploration problems [in Machine intelligence 10. J. E. Hayes et aI. (Eds.), John Wiley & Sons, Inc., New York, 1982,301-323. See main entry CR 24, 2 (Feb. 1983), Rev. 40,044.]
|
 |
|
Post a Comment
|
|
|
|
|
CONTRIBUTORS:
|
|
|
JOURNAL:
|
Computing Reviews,
24(3),
144 -
145.
|
|
|
|
YEAR:
|
1983
|
|
PUB TYPE:
|
Book Review
|
|
SUBJECT(S):
|
ARTIFICIAL INTELLIGENCE: Applications and Expert Systems MEASUREMENT. PERFORMANCE
|
|
DISCIPLINE:
|
Computer Science
|
|
HTTP:
|
|
|
LANGUAGE:
|
English
|
|
PUB ID:
|
103-426-858
(Last edited on
2006/05/24 12:33:07 GMT-6)
|
|
ABSTRACT:
PROSPECTOR is a computer program which can be used by geologists to determine mineral deposits; particularly in remote geographic locations where personal access is difficult or impossible. It is described by the author as a consultation system, since geologists tend to be apprehensive regarding problem solving by computer. In fact, it is stated that "most geologists have little or no experience with computers. . . [but] seem to be quite comfortable with questionnaires. . ." The thrust of the chapter is to validate PROSPECTOR by comparing program output with human geologist output. If the program is able to predict a mineral deposit with the same reasonableness as the professional geologist, then it would be a high-fidelity prognosticator.
The PROSPECTOR program includes knowledge about a particular type of ore deposit which is encoded in a flowchart/mathematical model representing observable geological features and relative quantitative values of these features. RWSSU represents a class of Western States sandstone uranium deposits and appears in several diagrams depicting the inference network. Since the complete RWSSU flowchart model spans 31 pages and appears in an earlier publication, only a small specimen is shown in the paper.
One type of validation performed compares the conclusions of PROSPECTOR with the independent conclusions of the expert geologist who designed the program. The geologist completes a questionnaire; comparisons rely on a - 5 to + 5 scale yielding numerical differences which are then evaluated. One feature of PROSPECTOR is the ability to explain its conclusions at any desired level of detail. The validation process resulted in an acceptable performance level. The author also describes the method of applying sensitivity analysis to the process.
Satisfied with the PROSPECTOR performance, a pilot was conducted to evaluate five regions selected by the US Department of Energy. Both US Geological Survey and DoE geologists were given questionnaires to obtain expert conclusions for comparison. There was generalIy close agreement with PROSPECTOR output, reflecting the program's capability to "synthesize many diverse factors, mechanically ascertaining general commonalities without being unduly distracted by occasional disparities." In ranking the test regions, the comparisons with expert predictions were very favorable in three areas, moderately favorable in one, and produced no match in one.
PROSPECTOR was also employed to evaluate several regions on the Alaskan Peninsula for uranium potential, and to quantitatively measure the economic value of a geological map for USGS, involving a porphyry copper model in PROSPECTOR. During the validation process, PROSPECTOR made its first prediction about the location of an as yet undiscovered ore body! This involved porphyry molybdenum deposits, but the display or output methodology differs from those described earlier. The geologic map is gridded into cells each 30 meters square.
Using data obtained by digitizing map features, PROSPECTOR evaluates each cell, using an efficient network compiler, and then color-codes and displays 16,384 cells on
a CRT.
PROSPECTOR represents an exciting area for continuing activity. It has great importance for economic geology, particularly for regions which are difficult for field geologists to reach. When the models are completed which repre~ent the logical process of the human geologist, then programs such as PROSPECTOR can move forward to generate new models which go beyond present geologist logic! At that point, geology will learn from computer systems rather than the reverse. It is that aspect of this effort at the Artificial Intelligence Center of SRI which is so valuable. In generations to come, geologists may be persuaded to make the leap from paper questionnaires and checklists to programs such as the one described in this chapter.
-L. C. Silvern, Sedona, AZ
|
|
|
|
STATISTICS
|
|
Click on # to view
|
|
Citations
|
|
1
|
|
References
|
|
0
|
|
Comments
|
|
0
|
|
Quality
|
|
0/0.00
|
|
Interest
|
|
0/0.00
|
|
View(er)s
|
|
2/81
|
|
|
|
|
|
|
| Prev |
Next |
|