Neuro-fuzzy modeling based genetic algorithms for identification of geochemical anomalies in mining geochemistry

Abstract

Abstract A genetic algorithm (GA)-based neuro-fuzzy approach is used for identification of geochemical anomalies by implementing a Takagi, Sugeno and Kang (TSK) type fuzzy inference system in a 5-layered feed-forward adaptive artificial neural network. This paper investigates the effectiveness of GA-based neuro-fuzzy for separating zone dispersed mineralization (ZDM) from blind mineralization, and its application for identification of geochemical anomalies in the arid landscape of the Lut metallogenic province in eastern Iran. Other classification algorithms such as metallometry, zonality, criteria, and back-propagation artificial neural network classifiers are also used for comparison. The genetic operators are carefully designed to optimize the artificial neural network, avoiding premature convergence and permutation problems. The results show that the GA-based hybrid neuro-fuzzy model can provide accurate results in comparison with those results obtained by other techniques. Neuro-fuzzy and GA-based neuro-fuzzy techniques appear to be well-suited for routine exploration geochemistry applications. In conjunction with statistics and conventional mathematical methods, hybrid approaches can be developed and may prove a step forward in the practice of applied geochemistry.