Mineral potential mapping using feed-forward neural networks

Mineral potential mapping is the process of producing a map that depicts the favorability of mineralization occurring over a specified region. The map should reflect the location of known mineral occurrences and also predict the distribution of areas of high mineral potential where little or no mining activity currently exists. Although the development of geographic information system technology and digital data manipulation techniques has enabled mineral exploration geologists to make more efficient use of resource information, many of the methods used are still inherently based on traditional techniques of map stacking in which layers of data are combined under the guidance of a mineral deposition model. This paper describes a data-driven mineral potential mapping technique based on feed-forward neural networks. Results are provided from applying the technique to gold exploration in a region of South West Victoria, Australia, using a range of geological, geophysical and geochemical input variables.