“…There is a variety of statistical methods for mineral prospectivity mapping that can be categorized into data-driven, knowledge-driven, and hybrids of them. Data-driven methods are the most popular mineral prospectivity mapping techniques, which are often theoretically based on regression/classification algorithms that come from machine learning field, for example, feed-forward neural networks (Brown et al, 2000(Brown et al, , 2003aHarris and Pan, 1999;Harris et al, 2003;Oh and Lee, 2010;Skabar, 2003), multilayer perceptrons (Chen, 2015;Skabar, 2005Skabar, , 2007, Bayesian networks (Porwal et al, 2006), radial basis functional link net (Behnia, 2007;Leite et al, 2009a;Nykänen, 2008;Porwal et al, 2003), probabilistic neural networks (Leite et al, 2009b), support vector machines (Abedi et al, 2012;Geranian et al, 2016;Zuo and Carranza, 2011), and random forests Laborte, 2015a, b, 2016;McKay and Harris, 2016; A C C E P T E D M A N U S C R I P T ACCEPTED MANUSCRIPT 6 mineral prospectivity, and (d) the geological features that are critical for polymetallic mineralization and can be used as spatial recognition criterion of polymetallic prospectivity in the study area.…”