105
and SVM had significantly better results than k-means, probably due to k-means not using
training data. This indicated that for this type of data it is better to use a supervised or
semi-supervised method. In Test 2 the highest accuracies were for shale, as BIF and ore are
easier to misclassify due to their higher similarity. Test 3 had higher accuracies than Test 1,
indicating that geochemical assays are a better indicator of WA or MN. Although they had
similar results, autoencoder is preferable to SVM for this application as it can handle more
than two categories. Overall, deep learning using autoencoder produced a high accuracy and
is an effective machine learning technique to use when to classifying this type of geological
and geochemical data.
REFERENCES
Balamurali, M. & Melkumyan, A. 2016. “t-SNE based visualisation and clustering of geological
domain,” in International Conference on Neural Information Processing. Springer, 2016, pp. 565–
572. In Hirose, A., Ozawa, S., Doya, K., Ikeda, K., Lee, M. & Liu, D. (eds.) Neural Information
Processing. ICONIP 2016. Lecture Notes in Computer Science. Springer, Cham.
Bengio, Y. 2009. Learning Deep Architectures for AI. Foundations and Trends in Machine Learning 2.
Clout, J.M.F. 2006. Iron formation-hosted iron ores in the Hamersley Province of Western Australia.
Transactions of the Institution of Mining and Metallurgy Section B, 115: 115–125.
Lascelles, D.F. 2000. Marra Mamba iron formation stratigraphy in the eastern Chichester range, Western Australia. Australian Journal of Earth Sciences: 799–806.
Liu, P., Choo, K., Wang, L. & Huang, F. 2016. SVM or deep learning? A comparative study on remote
sensing image classification. Soft Computing, 21: 7053–7065.
Melgani, F. & Bruzzone, L. 2004. Classification of hyperspectral remote sensing images with support
vector machines. IEEE Trans Geosci Remote Sens, 42: 1778–1790.
Nathan, D., Duuring, P., Holden, E.J., Wedge, D. & Horrocks, T. 2017. Learning characteristic natural
gamma shale marker signatures in iron ore deposits. Computers & Geosciences, 106: 77–88.
Silversides, K.L. & Melkumyan, A. 2016. Gaussian Processes based fusion of multiple data sources
for automatic identification of geological boundaries in mining. In Hirose, A., Ozawa, S., Doya, K.,
Ikeda, K., Lee, M. & Liu, D. (eds.) Neural Information Processing. ICONIP 2016. Lecture Notes in
Computer Science. Springer, Cham.
Steinhaus, H. 1957. Sur la division des corps matériels en parties. Bull. Acad. Sci. (in French), 4:
801–804.
Wamex A95838 2012. Wee One Annual Technical Report. http://geodocs.dmp.wa.gov.au/search.
jsp?Report_Ref=A95838&cabinetId=2301.
and SVM had significantly better results than k-means, probably due to k-means not using
training data. This indicated that for this type of data it is better to use a supervised or
semi-supervised method. In Test 2 the highest accuracies were for shale, as BIF and ore are
easier to misclassify due to their higher similarity. Test 3 had higher accuracies than Test 1,
indicating that geochemical assays are a better indicator of WA or MN. Although they had
similar results, autoencoder is preferable to SVM for this application as it can handle more
than two categories. Overall, deep learning using autoencoder produced a high accuracy and
is an effective machine learning technique to use when to classifying this type of geological
and geochemical data.
REFERENCES
Balamurali, M. & Melkumyan, A. 2016. “t-SNE based visualisation and clustering of geological
domain,” in International Conference on Neural Information Processing. Springer, 2016, pp. 565–
572. In Hirose, A., Ozawa, S., Doya, K., Ikeda, K., Lee, M. & Liu, D. (eds.) Neural Information
Processing. ICONIP 2016. Lecture Notes in Computer Science. Springer, Cham.
Bengio, Y. 2009. Learning Deep Architectures for AI. Foundations and Trends in Machine Learning 2.
Clout, J.M.F. 2006. Iron formation-hosted iron ores in the Hamersley Province of Western Australia.
Transactions of the Institution of Mining and Metallurgy Section B, 115: 115–125.
Lascelles, D.F. 2000. Marra Mamba iron formation stratigraphy in the eastern Chichester range, Western Australia. Australian Journal of Earth Sciences: 799–806.
Liu, P., Choo, K., Wang, L. & Huang, F. 2016. SVM or deep learning? A comparative study on remote
sensing image classification. Soft Computing, 21: 7053–7065.
Melgani, F. & Bruzzone, L. 2004. Classification of hyperspectral remote sensing images with support
vector machines. IEEE Trans Geosci Remote Sens, 42: 1778–1790.
Nathan, D., Duuring, P., Holden, E.J., Wedge, D. & Horrocks, T. 2017. Learning characteristic natural
gamma shale marker signatures in iron ore deposits. Computers & Geosciences, 106: 77–88.
Silversides, K.L. & Melkumyan, A. 2016. Gaussian Processes based fusion of multiple data sources
for automatic identification of geological boundaries in mining. In Hirose, A., Ozawa, S., Doya, K.,
Ikeda, K., Lee, M. & Liu, D. (eds.) Neural Information Processing. ICONIP 2016. Lecture Notes in
Computer Science. Springer, Cham.
Steinhaus, H. 1957. Sur la division des corps matériels en parties. Bull. Acad. Sci. (in French), 4:
801–804.
Wamex A95838 2012. Wee One Annual Technical Report. http://geodocs.dmp.wa.gov.au/search.
jsp?Report_Ref=A95838&cabinetId=2301.