Revisão do uso de técnicas de agrupamento para definição de domínios estacionários

Abstract

The definition of stationary domains is one of the first steps in mineral resource modeling. The incorrect grouping of samples can compromise the subsequent steps of modeling and even the estimation results, generating greater uncertainties in masses and grade values. The definition of stationary domains is most often confused with geological domains, which is not only subjective, but it also does not consider the correlations of the samples on multivariate or geographical spaces. This monograph aims to provide a wide bibliographical review about cluster algorithms which present interesting results and contribute for better stationary interpretation of the geostatistical data set and its validation. From traditional grouping techniques for statistical data – such as hierarchical agglomeration algorithm and k-means – to more recent techniques of spatial clusters that consider geographic positions of the samples – geostatistical hierarchical algorithm and double space agglomeration algorithm – all are discussed. As much as spatial algorithms have more elegant applicability and support in geostatistical data, a comparison with the results of traditional algorithms is necessary for comparative purposes, since validation is still a measure that depends on the knowledge of the geomodeler. Once different results and validations of the algorithms are compared, the geomodeler will have more grounding in deciding the most appropriate stationary domains. As laborious as the process can be, the application of these algorithms ensures that the next steps of resource modeling are not compromised, thus avoiding rework or even significant errors in the final estimate.