Ore deposits are the end product of a series of complex geological processes that operate over time and scales. Given the importance of the time- and scale-dependent processes, this study aims to develop a mineral prospectivity modeling method through contribution of the chronology of ore deposition processes. To achieve this goal, three different architectures of recurrent neural networks (RNNs), i.e., simpleRNN (SRNN), long short-term memory (LSTM), and gated recurrent unit (GRU), were examined to integrate layers of mineral system-based exploration criteria for prospectivity mapping. To compare the time sequence-based prospectivity modeling method (TMPM), which was generated using RNNs, with existing MPM approaches that don't consider the sequence of the ore-forming geological events in the modeling procedure, we generated two prospectivity models using convolutional neural network (CNN) and a classical fuzzy gamma operator. The results obtained demonstrated excellent performance of the three RNN methods over the CNN and fuzzy approaches. To illustrate and demonstrate the method proposed we used a data set of Mississippi Valley-type (MVT) Pb–Zn mineralization in the west of Semnan province, Iran.
