Accuracy assessment of various supervised machine learning algorithms in litho-facies classification from seismic data in the Penobscot field, Scotian Basin

Litho-facies classification is an essential task in characterizing the complex reservoirs petroleum exploration and subsequent field development. The lithofacies at borehole locations detailed but lacks providing larger coverage areas. acquired 3D seismic data provides global for studying reservoir facies heterogeneities study area. This applies six supervised machine learning techniques (Random Forest, Support Vector Machine, Artificial Neural Network, Adaptive Boosting, Xtreme Gradient Multilayer Perceptron) to post-stack accurately estimate different litho-facies inter-well regions compares their performance. Initially, efficacy of said models was critically examined via confusion matrix (accuracy misclass) evaluation (precision, recall, F1-score) on test data. It found that all performed best classifying shale (87%–94%) followed by sand (65%–79%) carbonate (60%–78%) Penobscot field, Scotian Basin. On overall accuracy scale, we multilayer perceptron method best-performing tool, whereas adaptive boosting least-performing tool three current analysis. While other methods also moderately good litho-facies. predicted using attributes matched well with log interpreted locations. indicates estimated are accurate reliable. Furthermore, validated results ascertain reliability between recommends applications reduce risk associated characterization.