 |
|
Kansas Geological Survey Open-file Report 2003-68 |
Accuracy statistics comparisons show the improvement in neural network predictions when geologic constraining variables are added. Absolute accuracy increases from 58% with no GCV's to 64% with M-NM and to 70% with RelPos added. Accuracy within one lithofacies and proportional representation show similar improvements.
GCV's additions were accomplished in the following manner:
- Marine-Nonmarine GCV extracted from Kansas Geological Survey oracle tops database and written to an LAS file
- Relative position curve generated in similar fashion
- GCV’s and E-Log variables combined into single LAS file.
- Neural network models pointed towards LAS file set, predictions made
and written to LAS file with nine curves (eight lithofacies probabilities
and one discrete lithofacies, the most probable)
    |
e-mail : webadmin@kgs.ku.edu
Last updated January 2004
http://www.kgs.ku.edu/PRS/publication/2003/ofr2003-68/slide15.html