Hybrid fuzzy c-means clustering algorithm and multilayer perceptron for increasing the estimate accuracy of the geochemical element concentration case study: eastern zone of porphyry copper deposit of Sonajil

Pattern recognition methods are able to identify the hidden relationships between exploration data, especially in the case of limited number of data. The geochemical distribution patterns of the elements are identified and generalized using these methods. Multilayer perceptron, MLP, is one of the pattern recognition methods which is used for the estimation of geochemical element concentrations in mineral deposit studies. In the current study, multilayer neural network was used to estimate the concentration of geochemical elements based on 1755 surface and borehole samples, analyzed by ICP. Fuzzy c-means, FCM, clustering algorithm was used to increase the neural network estimation accuracy. The optimal number of clusters in the dataset was identified by validation indices and was used to design estimator. The clustering data on average showed an increase of 13% accuracy compared to normal mode. The average accuracy was increased from 75 percent to 88 percent. Elements with the lowest estimation accuracy showed an acceptable increase on the estimation accuracy by using clustering data. Mean squared error was 0.079 using all data and decreased to 0.025 while using hybrid developed method.