Page 45 - Machine Learning for Subsurface Characterization

30 Machine learning for subsurface characterization

TABLE 1.2 Performances of the four unsupervised ODTs on Dataset #2

Dataset #2 results
Balanced accuracy F1 ROC-AUC
score score score
Isolation forest FS1 0.93 0.23 0.97
FS2 0.64 0.11 0.87
FS2** 0.86 0.21 0.92
FS3 0.91 0.22 0.96
FS4 0.93 0.24 0.99
One-class SVM FS1 0.76 0.22 0.89
FS2 0.6 0.11 0.87
FS2** 0.65 0.14 0.88
FS3 0.74 0.21 0.88
FS4 0.84 0.28 0.95
Local outlier FS1 0.38 0.11 0.62
factor
FS2 0.57 0.07 0.61
FS2** 0.56 0.08 0.63
FS3 0.61 0.1 0.62
FS4 0.61 0.09 0.55
DBSCAN FS1 0.58 0.18 NA
FS2 0.53 0.09 NA
FS2** 0.56 0.17 NA
FS3 0.58 0.14 NA
FS4 0.61 0.18 NA
Visual representation of the performances in terms of balanced accuracy score is shown in Fig. 1.7B.
**RT is replaced by RXO.

5.4 Performance on Dataset #4 containing manually labeled outliers
The offshore dataset contains seven log responses from different lithology,
namely, limestones, sandstone, dolomite, shale, and anhydrites. The seven logs
are gamma ray (GR), density (DEN), neutron porosity (NEU), compressional
sonic transit time (AC), deep and medium resistivities (RDEP and RMED),