3D mineral prospectivity modeling based on machine learning: A case study of the Zhuxi tungsten deposit in northeastern Jiangxi Province, South China

Abstract The Zhuxi tungsten deposit is the largest tungsten deposit in the world. A mineral prospectivity analysis of the areas surrounding this deposit has the potential to identify further resources, and thus, it is important to conduct such an analysis. Therefore, a mineral prospectivity was performed on the results of completed 3D geological modeling of the area and 3D inversion of gravity, magnetic, magnetotelluric and broadband seismic data. The anomalous density, susceptibility, resistivity, P-wave velocity and lithology derived from geophysical and geological modeling were constructed on consistent grids. Then, based on the locations of known deposits, the features of the ore-bearing and ore-free zones were extracted. Then, three kinds of machine learning algorithms, namely K-nearest neighbour, back-propagation neural network and support vector machine, were trained using the extracted features. Then, prediction was applied to the 3D data for the whole area, and multiple groups of 3D mineral prospectivity modeling were obtained. By a comparison of stage percentage and the analysis of the capture efficiency diagram, the support vector machine model without lithology captures the largest number of known ore-bearing samples (75.69%) with the least voxels (2.07%), which was highly consistent with the actual geological conditions selected for the classification of prospective areas. The results show that this type of machine learning-based mineral prospectivity modeling, which uses Earth models obtained from the 3D inversion of multiple geophysical data sets and 3D geological modeling, can advance 3D mineral prediction, and thus will greatly improve the efficiency in discovering further ore deposits.