Identification of Porphyry Copper Anomalies Using Support Vector Regression

Abstract

In the present study, data obtained from 83 rock samples collected from a porphyry copper mineralization zone in Canada were analyzed. First, the F-test feature selection method was employed to identify and determine the significant features, whereby five elements, gold (Au), molybdenum (Mo), silver (Ag), iron (Fe), and lead (Pb), were selected as the most important features for implementing the regression models. Subsequently, 70% of the data were allocated for training, and the remaining data were used for testing the models. To predict copper values, several methods, including the linear regression model, decision tree regression model, bagging ensemble tree model, boosting ensemble tree model, and a three-layer neural network model, were applied. The data were then predicted using the novel Projection Support Vector Regression (PSVR) model, and the Root Mean Square Error (RMSE), R-squared (R²), and training time of each method were analyzed. Furthermore, maps of prospective mineralized zones were generated in the Geographic Information System (GIS) environment based on the predicted copper values using the aforementioned methods. Finally, the results of all methods were compared, demonstrating the efficiency and superiority of the proposed PSVR method over the other investigated approaches.