| Abstract Submission No. | ABS-4.1-0073 |
| Title of Abstract | Combining machine learning and geostatistics for supervised classification of regionalized data |
| Authors | Abhishek Borah* |
| Organisation | Universidad de Chile |
| Address | Nabapur, Mazgaon Tezpur, Assam, India Pincode: 784001 E-mail: abhishekborah0@gmail.com |
| Country | India |
| Presentation | Poster |
| Abstract | The task of predicting the correct class label of a categorical response in a spatial context arises in many disciplines of the natural sciences, social sciences, and engineering. State-of-the-art techniques to typify and quantify hydrothermal alteration include the application of mineralogical (e.g., alteration indices that use normative minerals) and chemical (e.g., mass balance calculations) approaches, which have their inherent disadvantages such as sensitivity to the composition of the precursors, difficulty of reconciling with mineralogical observations, and a component of subjectivity. Another pathway is to use supervised classification via Machine Learning (ML) based on geochemical analyses of drill cores. However, such a classification task faces specific difficulties as ML classification is often designed for independent observations, while geochemical analyses are spatially autocorrelated. In this study, we incorporate geostatistics and machine learning for supervised classification of multivariate regionalized data, with an application to hydrothermal alteration mapping in a porphyry Cu-Au deposit (Mongolia). Among the many directions in the literature, we follow the approach by Dutta and Emery (2024), where geostatistics is used to replace the "noisy" feature variables by "denoised proxies", as an input for machine learning classification. The novelty is the application of geostatistical simulation instead of kriging to create the proxies and the fact that the proxies are processed differently than Adeli and Guartan (2021), who train as many classifiers as realizations. In a nutshell, we demonstrate two things in this study: the first is how geostatistics enriches machine learning to achieve higher predictive performance and to handle incomplete and noisy datasets in a spatial setting. Secondly, it establishes that a better prediction accuracy can be achieved than in previous studies where alteration types were predicted from geochemical data alone. |
| Keywords | Geostatistics, Machine Learning, Regionalized Data, Hydrothermal Alteration, Simulation, Mining |