Catalog Home Page

Applying complementary neural networks to porosity prediction in well log data analysis

Kraipeerapun, P., Fung, C.C., Nakkrasae, S. and Amornsamankul, S. (2009) Applying complementary neural networks to porosity prediction in well log data analysis. In: 6th International Joint Conference on Computer Science and Software Engineering (JCSSE2009), 13 - 15 May, Phuket, Thailand

[img]
Preview

Abstract

This paper presents a novel approach to the regression problem using complementary neural networks (CMTNN). The experiment is realized to the problem of porosity prediction in well log data analysis. Two sets of complementary neural networks are created. The first set is a pair of opposite neural networks trained to predict degree of truth porosity and degree of falsity porosity values. The second set is also a pair of opposite neural networks trained with the same input and parameters as the first set; however, they are trained to predict degree of truth gap values and degree of falsity gap values. The gap is the difference between the truth and falsity porosity values. The predicted truth and falsity gap values are combined and considered as the predicted gap. After that, the predicted gap is compared to the gap which is the difference between the predicted truth and falsity porosity values. The difference between both gaps is considered as an uncertainty value. Moreover, uncertainty of type error occurred in the prediction is also considered in this study. In this paper, the aggregation based on dynamic weight combination is proposed. Uncertainty values are used as weights associated with the truth and falsity porosity values. The results obtained from our approach are compared to results obtained from traditional models that use only the truth porosity values in the prediction. These existing models are based on backpropagation neural network and support vector regression. We found that our approach improves performance compared to those existing models.

Publication Type: Conference Paper
Murdoch Affiliation: School of Information Technology
URI: http://researchrepository.murdoch.edu.au/id/eprint/21641
Item Control Page Item Control Page

Downloads

Downloads per month over past year