Abstract:

Quantitative prediction of mineral resources needs the support of data science urgently as the field has now changed from qualitative to quantitative, from data sparse to data intensive. On the basis of previous studies, this paper discusses data science-based theory and method of quantitative prediction of mineral resources. The theoretical basis of such theory and method are correlation theory and anomaly theory. The former, via supervised machine learning algorithms, provides a theoretical basis for the prediction of undiscovered mineral deposits by mining the correlations between geological prospecting big data and locations of mineral deposits; the latter, by detecting geological anomaly present in geological prospecting big data, provides a theoretical basis for the prediction of mineral deposits. This data science-based approach emphasizes the importance of geological prospecting big data and machine learning algorithms, as the type, diversity, quality and accuracy of geospatial data can affect the final prediction results, whilst machine learning algorithms can improve the efficiency of feature extraction and information integration fusion. This paper presents the workflow of quantitative prediction of mineral resources by the data science-based theory and method, introduces the methods for feature extraction and prospecting information fusion, and discusses potential prediction uncertainty inherent in such theory and method.

Key words: data science, quantitative prediction of mineral resources, geological prospecting big data, machine learning, uncertainty