Optimal Selection of Classification Algorithms for Well Log Interpretation
Guangren Shi, Jinshan Ma, Dan Ba · 2015
Three classification algorithms and one regression algorithm have been applied to well log interpretation. The three classification algorithms are the classification of support vector machine (C-SVM), the naive Bayesian (NBAY), and the Bayesian successive discrimination (BAYSD), while the one regression algorithm is the multiple regression analysis (MRA). In these four algorithms, only MRA is a linear algorithm whereas the other three are nonlinear algorithms. In general, when all these four algorithms are used to solve a real-world problem, they often produce different solution accuracies. Toward this issue, the solution accuracy is expressed with the total mean absolute relative residual for all samples, R(%). Then three criteria have been proposed: 1) nonlinearity degree of a studied problem based on R(%) of MRA (weak if R(%)<10, and strong if R(%)≥10); 2) solution accuracy of a given algorithm application based on its R(%) (high if R(%)<10, and low if R(%)≥10); and 3) results availability of a given algorithm application based on its R(%) (applicable if R(%)<10, and inapplicable if R(%)≥10). Four case studies have been used to validate the proposed approach. These four case studies are a classification problem. The calculation results indicate that a) when a case study is a weakly nonlinear problem, C-SVM, NBAY and BAYSD are all applicable, and BAYSD is better than CSVM and NBAY; b) when a case study is a strongly nonlinear problem, C-SVM is applicable, NBAY is inapplicable, whereas BAYSD is sometimes applicable; and c) BAYSD and C-SVM can be applied to dimensionality reduction.