Unsupervised Machine Learning for Regionalization of Environmental Data: Distribution of Uranium in Groundwater in Ukraine
ISBN 978-85-88783-11-9
Authors
1Govorov, M.; 2Gienko, G.; 3Putrenko, V.
1VANCOUVER ISLAND UNIVERSITY Email: govorovm@viu.ca
2UNIVERSITY OF ALASKA ANCHORAGE Email: ggienko@uaa.alaska.edu
3NATIONAL TECHNICAL UNIVERSITY OF UKRAINE Email: putrenko@rambler.ru
Abstract
In this paper, several unsupervised machine learning algorithms were explored to define homogeneous regions of concentration of uranium in surface waters using multiple environmental parameters. Primary envi-ronmental parameters were identified using several spatial statistical methods explored in the previous study. At this step, cluster analysis was carried out using techniques of bivariate local pattern analysis, spatially contiguous clustering of multivariate data or unsupervised learning, and techniques from the domain of artificial neural network. Combining techniques of data-based and model-based classifications of geological, climatic, and other environmental indicators coupled with spatial statistical analysis allowed to identify six regions of distribution of natural radioactive elements in groundwater in Ukraine.