Porphyry copper prospectivity modelling using data driven and hybrid outranking methods: A case study of Shahr-e-Babak study area, South Eastern Iran

Abstract

Identifying potential and mineralized areas with a reasonable level of confidence is a complex issue. The use of supervised methods and machine learning could help to achieve a batter results in such studies. In the present study, exploratory data and information available in the study area of Shahr-e-Babak were collected, pre-processed, processed and analyzed. In the next step, this information was implemented to produce and prioritize a porphyry copper mineralization model. In order to improve the modelling of porphyry copper mineralization, a combination of data-driven methods, supervised machine learning, and multivariate decision-making (MCDM) methods were applied. This paper proposes an acceptable process for selecting exploration layers, the impact of weights on the generated layers, and their combination. For this purpose, a data-driven method was used to select exploratory evidential layers. Then, machine learning methods include random forest (RF), adaptive neuro fuzzy (ANFIS), artificial neural network (ANN) and generalized neural network (GRNN), which are proven methods in mineral prospectivity modelling (MPM), were used to integrate exploration layers. The exploratory evidential layers include remote sensing, geochemistry, geology, and geophysics. To improve the obtained models, decrease stochastic uncertainty, prioritize porphyry copper potential area and generate the final MPM, the MOORA method as a MCDM method was used to. Then, the prediction-area plot (P-A) method, taking into account the cu occurrences in the area and the normalized density method as a traditional method, were applied to weight and evaluate the produced layers in the form of a data-driven method. Subsequently, the final potential map was generated using the MOORA method with more favourable performance than machine learning methods. The results of the MOORA method confirm that this process is more successful in producing the desired MPM.