Unsupervised Machine Learning for Electrofacies Classification in Offshore Keta Basin

A recent study introduces an unsupervised machine learning approach for analyzing electrofacies in Ghana's offshore Keta Basin, where core data is limited. Researchers examined six standard wireline logs from Well C, covering a depth range of about 11,195 samples. Using K-means clustering in a multivariate log space, they assessed the clustering structure through inertia and silhouette diagnostics. The analysis revealed four distinct clusters, with an average silhouette coefficient near 0.50, indicating a moderate yet significant separation. The electrofacies identified display consistent depth-related patterns linked to changes in clay content, porosity, and rock framework properties, illustrating a geological transition from shale-dominated to cleaner sandstone-dominated units. This study highlights that log-only, unsupervised clustering, backed by quantitative metrics, offers a reliable framework for electrofacies classification and porosity analysis in environments with limited data.