Today’s explorationists have unprecedented capacity to acquire new and historic geological, geochemical and geophysical data; however, the integrated analysis and interpretation of such data remains a significant challenge. A computational approach based on self-organising maps (SOM) can assist in understanding and synthesizing such data. A SOM analysis can highlight subtle relationships and assist in the process of knowledge creation from complex and disparate data. In a SOM analysis each sample is treated as a vector in a data space determined by its variables; and, measures of vector similarity are used to order and segment the input data into meaningful natural patterns. Because SOM is an exploratory data analysis tool that is unsupervised and data-driven, the resulting patterns, boundaries and relationships are internally-derived. Typically, one of the primary SOM outputs is a rectilinear map. This map is an orderly two-dimensional representation of the multidimensional input data set, which displays relationships between the input samples. More importantly, the map can also be used as a framework to display the variables’ contributions related to those samples. By judicious application of some form of colour look-up table, it is possible to assess the spatial coherence or context of samples belonging to a particular node, or group of nodes (in the assigned colour of that node). Furthermore, scatterplots of variable values belonging to nodes (also coloured by the node look-up table) are an effective means of identifying subtle trends and relationships such as identifying and separating overlapping geological processes, related to mineralization and subsequent weathering or metamorphic events. As our technological capacity to acquire data increases, evidence-based knowledge extraction techniques will become increasingly important. Because of its vector-quantization approach and its ability to analyze, integrate and allow an integrated interpretation of complex and disparate data, SOM is an ideal tool to assist in this process.
[1]
Brian S. Penn,et al.
Using self-organizing maps to visualize high-dimensional data
,
2005,
Comput. Geosci..
[2]
Santiago Torres-Alegre,et al.
Using self-organizing maps to identify potential halo white dwarfs
,
2003,
Neural Networks.
[3]
Samuel Kaski,et al.
Bibliography of Self-Organizing Map (SOM) Papers: 1981-1997
,
1998
.
[4]
Traitement des diagraphies à l'aide d'un réseau de neurones du type « carte auto-organisatrice » : application à l'étude lithologique de la couche silteuse de Marcoule (Gard, France)
,
2002
.
[5]
Binzhong Zhou,et al.
New Approaches for Rock Strength Estimation from Geophysical Logs
,
2005
.
[6]
S. J. Fraser,et al.
Data mining mining data: ordered vector quantisation and examples of its application to mine geotechnical data sets
,
2006
.
[7]
B. Dickson,et al.
Quietening the noise: an evaluation of noise reduction methods applied to aerial gamma-ray survey data
,
2003
.
[8]
UWE STRECKER,et al.
Data mining of 3 D poststack seismic attribute volumes using Kohonen self-organizing maps
,
2002
.
[9]
Bruce Dickson,et al.
And now for "SOM" thing completely different: Spatial Data Mining
,
2001
.