Unsupervised Machine Learning Applications for Seismic Facies Classification

Unsupervised ML uses the attributes themselves as both training data and data to be analyzed. The simplest algorithm is K-means, wherein the interpreter defines the number of facies (clusters) to be found. The algorithm then finds means and standard deviations (more generally, covariance matrices) to determine the center and the extent of each cluster in multidimensional attribute space, and thus generates different clusters.