Prospectivity analysis using unsupervised machine learning

Prospectivity modeling involves integration of geological features that serve as spatial proxies of mineralization processes. These features are hand-crafted from primary geoscience datasets, which is time-consuming and labour-intensive. Also, the feature selection process is guided by conceptual modeling of the targeted deposit type, which may be biased by the subjective preference of the expert geologist. This study applies Self-Organising Maps (SOM), an unsupervised machine learning algorithm, to gridded geophysical and topographical datasets in order to identify and demarcate exploration targets for carbonatite-alkaline-complex-hosted REE deposits in the western Rajasthan in northwest India. Interpreted or manually generated data, such as surface or bed-rock geological maps, were not used in the study. The results are compared with those obtained from a previous supervised knowledge-driven prospectivity analysis. It was found that the results are comparable. Therefore, unsupervised machine learning algorithms are robust alternatives to knowledge-driven or supervised data-driven prospectivity modeling, particularly in unexplored terrains for which there is little or no geological knowledge available.