26th International Mining Congress and Exhibition of Turkey, IMCET 2019, Antalya, Türkiye, 16 - 19 Nisan 2019, ss.525-536
Copyright © IMCET 2019 International Mining Congress and Exhibition of Turkey. All rights reserved.Exploratory data analysis (EDA) has vital importance in resource estimation which involves improving the understanding of the domains. Sound geological interpretation of domains are the foundation for reliable variogram analysis - the basis of geostatistical estimation. This implies no mixing of populations occurs, as this violates the assumption of stationarity necessary for geostatistical estimation. EDA can be mainly divided into 2 groups: univariate and multivariate. The univariate representations produce representations of one variable while the multivarivate representations try to show the relationship between several variables. In this study, Self Organizing Maps which belong to a Non-Linear Dimensionality Reduction technique is presented in multidimensional rare earth elements (REE) database. This technique is typically used for visualization and interpretation, as well as variable selection in very large databases. Two REE datasets are used in this study. First, publicly accessible data that include major, minor, trace and REE concentration of geologic formations in Wyoming oil and gas fields. Second data is from REE deposit located in Turkey. Comprehensive SOM results are given in detail and interesting clustering schemes are discussed.