Regional-Scale Mineral Prospectivity Mapping: Support Vector Machines and an Improved Data-Driven Multi-criteria Decision-Making Technique

Abstract

Mapping mineral prospectivity (MPM) is mostly beset with prediction uncertainties, which are generally categorized into (a) stochastic and (b) systemic types. The stochastic type is usually linked to the low quality as well as insufficiency/inefficiency of data used. In contrast, inaccurate selection of exploration criteria, exaggerated and arbitrary weighting of spatial evidence layers resulting from subjective judgment of analyst and applying an integration methodology, which is not able to consider the complexities of geological processes, are main sources of systemic type. This paper aims for reducing the second type of MPM uncertainty in delineating favorable exploration targets for Cu-Au mineralization in the Moalleman District, NE Iran. Thus, several efficient evidence layers were translated from geospatial criteria (e.g., geochemical, geological, structural and hydrothermal alterations) and were considered for integration purpose in the study area. Then, an improved data-driven simple additive weight (data-driven SAW) procedure was introduced for generating prospectivity model. In this procedure, prediction-area plots and frequency ratio method were applied for assigning objective weights to efficient evidence layers and their corresponding classes, respectively. Furthermore, a supervised algorithm for machine learning classification namely support vector machine (SVM) with radial basis function kernel was executed for comparison purposes. The results indicated that the two prospectivity models are succeeded in delineating favorable targets of mineralization; however, the SVM model is more reliable than data-driven SAW in predicting high-potential areas of mineralization.