Unsupervised and supervised learning in cascade for petroleum geology

Cascade of unsupervised and supervised learning algorithms are suitable in all those problems where there are large unlabelled input datasets and the underlying data structure is hidden and not clearly defined. In petroleum geology the understanding and characterization of reservoirs needs integration of different subsurface data in order to create reliable reservoir models. The large amount of data for each well and the presence of different wells to be simultaneously analysed make this task both complex and time consuming. In this scenario, the development of reliable characterization methods is of prime importance in order to help the geologist and reduce the subjectivity of data interpretation. In this paper, we propose a novel interpretation system based on the use of unsupervised and supervised learning techniques in cascade. Using unsupervised algorithm the domain expert identifies relevant clusters that will be used as classes in the following step, in order to learn a classifier to be applied to new instances and wells. We test the approach over five real well dataset using different evaluating techniques. Main advantages of this approach are the ability to manage and use a large amount of data simultaneously and the reduction in interpretation time of a group of wells.