Neural Network Ensembles using Interval Neutrosophic Sets and Bagging for Mineral Prospectivity Prediction and Quantification of Uncertainty

This paper describes the integration of neural network ensembles and interval neutrosophic sets using bagging technique for predicting regional-scale potential for mineral deposits as well as quantifying uncertainty in the predictions. Uncertainty in the types of error and vagueness are considered in this paper. Each component in the ensemble consists of a pair of neural networks trained for predicting the degrees of favourability for deposit and barren. They are considered as the truth-membership and the false-membership values, respectively. Errors occurred in the prediction are estimated using multidimensional scaling and interpolation methods. Vagueness is computed as the difference between truthand false-membership values. In this study, uncertainty of type vagueness is determined as the indeterminacy-membership value. Together these three membership values form an interval neutrosophic set. In order to combine and classify outputs from components in the ensemble, three aggregation methods are proposed in this paper. Our proposed model improves the classification performance as compared to the simple majority vote and averaging methods