Full text loading...
-
Nearest Neighbor Based Seismic Attribute Selection for Effective Porosity Prediction
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 76th EAGE Conference and Exhibition 2014, Jun 2014, Volume 2014, p.1 - 5
Abstract
We applied a non-linear, robust, attribute selection method called RGS (Regression, Gradient guided, feature selection) to find an optimal subset of attributes for prediction of effective porosity in an oil field from south-west of Iran. This method utilizes a special kind of k-nearest-neighbor algorithm as predictor that makes attribute selection procedure fast and suitable for large scale problems. Comparing the results from this algorithm to common attribute selection methods indicated RGS is superior to other methods for reducing the curse of dimensionality by selecting lower number of attributes and resulting in the smallest validation error among all. Analyzing the results of this study indicated consideration of complex dependency of the target function on group of attributes helps RGS to achieve better performance.