Integration of multiscale attributes using a semisupervised machine learning algorithm
Salma Alsinan, Philippe Nivlet, Yazeed Altowairqi, Harald Karg · 2022
The paper presents a machine learning workflow to integrate data from basin modelling, rock physics, geochemistry, and production to predict regional sweet spots in unconventional reservoirs. The presented workflow uses a combination of different statistical learning techniques to reduce the dimensionality of the attributes and combine them into a joint qualitative label at well locations. A semi-supervised algorithm is then used to propagate the classified labels to the unlabeled data whilst adhering to a set of predefined spatial constraints. The results show the strength of data integration in reducing the uncertainty associated with each data type.