Análise Espacial Guiada Pelos Dados (Data-Driven): O Uso De Redes Neurais Para Avaliação Do Potencial Poli-Minerálico Na Região Centro-Leste Da Bahia

Targeting of Gold Deposits in Amazonian Exploration Frontiers using Knowledge- and Data-Driven Spatial Modeling of Geophysical, Geochemical, and Geological Data

“…In this work, the following methods were applied: weights-of-evidence (Agterberg et al 1990;Assadi and Hale 2000;Bonham-Carter et al 1989;Carranza and Hale 2000;Debba et al 2009;Roy et al 2006;Tangestani and Moore 2001), artificial neural networks (Bougrain et al 2003;Lammoglia et al 2007;Leite and Souza Filho 2009a, b;Nóbrega and Souza Filho 2003;Nykänen 2008;Porwal et al 2003), and fuzzy logic (Carranza and Hale 1997;Cheng and Agterberg 1999;Lee 2007;Nykänen et al 2008b;Quadros et al 2006).…”

“…The ANN method consists of an adaptive computational system that, by means of artificial intelligence, provides pattern recognition or data classification with the objective of conducting specific tasks, such as the spatial relationship between known gold occurrences and exploration data to produce mineral potential maps (Bougrain et al 2003;Leite and Souza Filho 2009a, b;Nóbrega and Souza Filho 2003;Porwal et al 2003). Some properties make the neural networks a technique suitable for pattern recognition and classification of spatial data: (1) the ability to extract hidden patterns that may not be perceptible to humans or to traditional statistical techniques; (2) the capacity to analyze data without the need of previous information or a mineral deposit model; (3) the possibility of working with noisy, limited, interdependent, or non-linear data; (4) the possibility of continuous addition of new data; and (5) the capacity to analyze large datasets (Brown et al 2000;Lammoglia et al 2007;Nóbrega and Souza Filho 2003).…”

“…The Fire Assay method was used to determine Au grades in the samples, with a detection limit of 0.01 g/ton. Considering that each stream sediment sample represents materials upstream and upslope of the location it was collected, gold values should not be interpolated but represented on the basis of sub-watersheds (Carranza and Hale 1997;Nóbrega and Souza Filho 2003). Using tools available in GIS, a digital terrain model (DTM) was used to generate a map of sub-watersheds (Jenson and Domingue 1988), according to the following steps: (1) drainage extraction and generation of a depression-free DTM, (2) generation of a matrix file with the flow direction of each cell, (3) generation of a matrix file of accumulated flow for each cell, (4) extraction of the drainage network considering an accumulated flow of 100 cells or larger, (5) conversion to a vector file, (6) delineation of the areas of the hydrographic watersheds and definition of values that corresponds to the watershed flow direction, for each drainage segment, and (7) generation of a sub-watershed map.…”

Lithofacies Recognition Based on Fuzzy Logic and Neural Networks: A Methodological Comparison

“…Neural networks are suitable for pattern recognition from datasets because of their (i) ability to extract hidden patterns that can be imperceptible to humans and other traditional statistical techniques; (ii) facility to analyze data without any prior knowledge about its distribution; (iii) possibility to work with noisy, limited, interdependent and non-linear data; (iv) performance and speed, particularly when spatial features have complex characteristics and sources comprise different statistical distributions; (v) option to add new data continuously as input layers; (vi) and capability to analyze large data sets (e.g., Hemilson & Crane, 1994;Brown et al, 2000;Nóbrega & Souza Filho, 2003).…”

“…These systems are adaptive and can learn or gain knowledge from data by means of artificial intelligence techniques (e.g., Nóbrega & Souza Filho, 2003). Generally, the network is composed of simple processing units (nodes or neurons).…”