A combined multivariate approach analyzing geochemical data for knowledge discovery: The Vazante – Paracatu Zinc District, Minas Gerais, Brazil

Cevik, Ilkay S.; Olivo, Gema R.; Ortíz, Julián M.

doi:10.1016/j.gexplo.2020.106696

Cited by 9 publications

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This means that t‐SNE aims to preserve the local relative sample distances in a given multi‐parameter data space (e.g., the radiogenic isotope ratios in oceanic basalts) with the aim to group data with similar characteristics together and visualize them in plots of two variables (2D data space). It has become a routine data analysis tool in the biological or computational sciences, but has rarely been applied in geochemistry, for example, in geochemical exploration (e.g., Balamurali and Melkumyan, 2016; Cevik et al., 2021; Horrocks et al., 2019). We show that t‐SNE is a powerful and robust machine learning technique for identifying global‐scale similarities between the individual, local data sets that contribute to the global MORB‐OIB radiogenic isotope database.…”

Section: Introductionmentioning

confidence: 99%

Chemical Geodynamics Insights From a Machine Learning Approach

Stracke

Willig

Genske

et al. 2022

Geochem Geophys Geosyst

View full text Add to dashboard Cite

Thousands of oceanic basalts have been analyzed in the last six decades, building up a large database of Sr, Nd, Hf, and Pb isotope ratios. A characteristic feature of these data is that they are not randomly distributed. In 2D and 3D isotope ratio diagrams, the isotopic variation of oceanic basalts on a local scale, that is, for basalts from one ocean island or ridge segment, is often in form of approximately linear data trends. Combining the local data into a large global data set does not have a randomizing effect. Rather, despite considerable overlap, higher-order structures, that is, reproducible trends or clusters composed of groups of local data, appear in 2-3D isotopic space. These higher-order trends or clusters of the global data set have often been described as linear combinations

show abstract

Section: Introductionmentioning

confidence: 99%