Clustering Seismic Datasets for Optimized Facies Analysis Using a SSCSOM Technique

The derivation of seismic facies codes needs extensive knowledge and selection of seismic attributes. Seismic facies analysis is aimed to find the natural grouping in a data set based on the character of the recorded seismic response. In a common unsupervised facies analysis algorithm it is found that each group of seismic trace segments are related to a particular depositional environment. Most of the previous attempts in seismic facies analysis were in application of unsupervised algorithms like SOM.Control in unsupervised clustering results is limited by the number of final clusters (facies codes) that is mostly fixed by predefined and existing lithofacies codes or wavelet shapes. The idea is having user-intervention and interpretation in the per class posterior probability space. Unlike supervised methods in which the labeling is done in the input space by picking, semi-supervised methods are aiming to accept/reject clustering on the basis of interpretation in the near well space.