A machine learning approach to tungsten prospectivity modelling using knowledge-driven feature extraction and model confidenc

作     者：Christopher M.Yeomans Robin K.Shail Stephen Grebby Vesa Nykanen Maarit Middleton Paul A.J.Lusty Christopher M. Yeomans;Robin K. Shail;Stephen Grebby;Vesa Nyk?nen;Maarit Middleton;Paul A.J. Lusty

作者机构：Camborne School of MinesCollege of EngineeringMathematics and Physical SciencesUniversity of ExeterPenryn CampusPenrynCornwallTR109FEUK British Geological SurveyEnvironmental Science CentreKeyworthNottinghamshireNG125GGUK University of NottinghamNottingham Geospatial InstituteInnovation ParkNottinghamNG72TUUK Geological Survey of FinlandP.O.Box 77FI-96101RovaniemiFinland 

出 版 物：《Geoscience Frontiers》 (地学前缘（英文版）)

年 卷 期：2020年第11卷第6期

页      面：2067-2081页

核心收录：

学科分类：081803[工学-地质工程] 08[工学] 0708[理学-地球物理学] 0818[工学-地质资源与地质工程] 0704[理学-天文学] 

基　　金：funded by the British Geological Survey,United Kingdom(S267) the Natural Environment Research Council(NERC),United Kingdom 

主　　题：Machine learning Mineral prospectivity modelling Mineral exploration Random ForestTM Tungsten SW England 

摘      要：Novel mineral prospectivity modelling presented here applies knowledge-driven feature extraction to a datadriven machine learning approach for tungsten *** method emphasises the importance of appropriate model evaluation and develops a new Confidence Metric to generate spatially refined and robust exploration *** data-driven Random ForestTM algorithm is employed to model tungsten mineralisation in SW England using a range of geological,geochemical and geophysical evidence layers which include a depth to granite evidence *** models are presented,one using standardised input variables and a second that implements fuzzy set theory as part of an augmented feature extraction *** use of fuzzy data transformations mean feature extraction can incorporate some user-knowledge about the mineralisation into the *** typically subjective approach is guided using the Receiver Operating Characteristics(ROC)curve tool where transformed data are compared to known training *** modelling is conducted using 34 known true positive samples with 10 sets of randomly generated true negative samples to test the random effect on the *** two models have similar accuracy but show different spatial distributions when identifying highly prospective *** analysis shows that the fuzzy-transformed model is a better discriminator and highlights three areas of high prospectivity that were not previously *** Confidence Metric,derived from model variance,is employed to further evaluate the *** new metric is useful for refining exploration targets and highlighting the most robust areas for follow-up *** fuzzy-transformed model is shown to contain larger areas of high model confidence compared to the model using standardised ***,legacy mining data,from drilling reports and mine descriptions,is used to further validate the fuzzy-transformed model and gauge the depth of potential *** of m