IEEE Geoscience and Remote Sensing Magazine - December 2015 - 32

find KNIME useful because of the memory management
capabilities, allowing to process on the order of billions of
records on a desktop PC with relatively small amount of
RAM (4 gigabytes).
Among the rare exceptions, in environmental science
KNIME finds application to the analysis of fire hazard in
forested areas [11]. The authors demonstrate that KNIME
can easily allow the use and comparison of multiple DM
methods to predict burned areas due to forest fires. In
geoscience we find numerous papers that apply similarly
DM to identify fossil fuels and discover mineral deposits [6], [7], [12], [13]. Feltrin et al. [14] apply a workflow
approach in KNIME to perform unsupervised learning
using k-means and fuzzy-c-means clustering, see also
[15]. This example is discussed in the next section with
the objective of demonstrating the tool applicability to
geoscience data.
6. A BRIEF EXAMPLE OF KNIME
KNIME examples are numerous as seen in the on-line
forum and technical documentation available on the
main web site (http://www.knime.org/). However, as discussed, most applications use chemical and biological
or other non-geological data. Thus, the following image
(Fig. 1) illustrates an example of a KNIME workflow
developed to classify geological domains in a selected
region, using two DM classification approaches, k-means
and fuzzy c-means clustering. In this example k-means
and fuzzy c-means clustering are linear classifiers used to
separate a multivariate space defined by two independent
multiclass continuous variables represented by geophysical survey data. The difference between the algorithms
is found in the way they consider class membership. The
k-means classification expresses full membership instead
fuzzy-c-means allows for partial degrees of membership
to a class/cluster family. In essence, the first methodology blindly assigns a sample to its closest cluster family,
whereas the second methodology assigns a sample to
either one or more than one cluster families introducing
an additional parameter that expresses the likelihood of
belonging to each class.
With a similar approach to [16] it is proposed a pseudolithology classification of a combination of airborne, Bouguer gravity data and reduced to the pole, total magnetic
intensity data based on chosen classification methods. The
KNIME workflow used to generate a pseudo-geology map
in [14] is composed of the following nodes (cf. Fig. 1 A): (1)
An import node used to convert ASCII files (geophysical
raster data) into a KNIME table; (2) pre-processing nodes
for the removal of missing information (raster images contained areas devoid of geophysical data) and data standardization (a requirement of cluster analysis is the normalization of input data); (3) a JFreeChart node (not shown in
Fig. 1) was used to visually check the results of data transformation; (4) two clustering nodes were used to classify
the combination of gravity and magnetic gridded data: a
32

k-means clustering [17], [18] and a Fuzzy-k-means clustering approach [12], [15].
Since clustering algorithms are influenced by outliers rather than assigning an arbitrary value to the MI,
MI rows were removed and added after processing as an
independent cluster in a post classification phase of the
analysis. (5) R nodes were used to reconstruct the raster
data providing a final output that was translated into
classified maps representative of pseudo-lithology (Fig. 1
B,C) These outputs can facilitate further geological interpretation [16]. The ad-vantage of this workflow based
approach derives from the capacity of recording each processing step which is useful for iterative experimentations
and debugging.
4
7. CONCLUSIVE REMARKS
With the popularization of tools such as KNIME we foresee an incremental expansion of the application of Data
Mining to various areas of geoscience where the number of
applications of ML and DM is still very limited [9]. KNIME
with its intuitive approach facilitates the application of
complex Machine Learning tools. Development of tailored
workflows in KNIME and custom-nodes to address dominantly data- but also knowledge-driven analysis in the geosciences may become an interesting area of computational
geoscience.
ACKNOWLEDGMENT
The author wishes to thank Dr. Martina Bertelli and anonymous reviewers for useful comments and proofreading of
the manuscript. Corresponding Author E-mail: lfeltrin@
uwo.ca; Fax: 1-519-661-3198.
REFERENCES
[1] M. R. Berthold, et al., "KNIME: the Konstanz information miner,"
SIGKDD Explor., vol. 11, no. 1, pp. 26-31, 2009.
[2] S. Beisken, et al., "KNIME-CDK: Workflow-driven cheminformatics.," BioMed Central Bioinform., vol. 14, p. 257, Aug. 2013.
[3] W. A. Warr, "Scientific workflow systems: Pipeline Pilot and
KNIME," J. Comput. Aided Mol. Des., vol. 26, no. 7, pp. 801-804,
2012.
[4] M. Berthold and D. J. Hand, Intelligent Data Analysis. Berlin, Germany: Springer, vol. 42, no. 7, 2010.
[5] E. C. Grunsky, "R: A data analysis and statistical programming
envi-ronment- an emerging tool for the geosciences," Comput.
Geosci., vol. 28, no. 10, pp. 1219-1222, 2002.
[6] M. Abedi, G. Norouzi, and A. Bahroudi, "Support vector machine
for multi-classification of mineral prospectivity areas," Comput.
Geosci., vol. 46, pp. 272-283, Sept. 2012.
[7] V. Rodriguez-Galiano, M. Sanchez-Castillo, M. Chica-Olmo, and
M. Chica-Rivas, "Machine learning predictive models for mineral prospectivity: An evaluation of neural networks, random
forest, regression trees and support vector machines," Ore Geol.
Rev., vol. 71, pp. 804-818, Dec. 2015.
[8] A. Zamani and N. Hashemi, "Computer-based self-organized tectonic zoning: A tentative pattern recognition for Iran," Comput.
Geosci., vol. 30, pp. 705-718, Aug. 2004.

(continued on p. 38)
ieee Geoscience and remote sensing magazine

december 2015

http://www.knime.org/

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2015