Chapter 1 - User vs. machine-based seismic attribute selection for unsupervised machine learning techniques: Does human insight provide better results than statistically chosen attributes?

Abstract

In geosciences, a variety of machine learning (ML) algorithms are currently being employed for multiple purposes, for example, facies classification, fault prediction, and reservoir characterization. Among these are two clustering methods: principal component analysis (PCA) and self-organized maps (SOMs), which provide a fast organization of data into groups or clusters (with no geologic supervision) that aid in preliminary geological interpretation. With increasingly common usage of these techniques, the motivation of this chapter is to investigate the impact of a user-controlled selection of attributes to perform SOM for deepwater seismic facies classification versus a machine-selected result through PCA. Results reveal that whereas an appropriate combination of attributes with a clear interpretation objective enhances the SOM’s results and facilitates the interpreter understanding of the output classes, PCA provides insightful information regarding the contribution of seismic attributes that may not have been initially considered. While machine learning techniques are a powerful “tool” for geological interpretation, user control on initial input attributes and validation of output using an “in-context” interpretation is necessary for an optimal elucidation, at least in unsupervised machine learning methods.