A Comparative Study Of The Performance Of Single Neural Network Vs. Adaboost Algorithm Based Combination Of Multiple Neural Networks For Mineral Resource Estimation

This paper investigates the performance of the neural network technique in a placer deposit situated at the Nome district in Alaska. Initially a single neural network model was constructed to estimate the gold grade. However, at a later stage, an ensemble model consisting of multiple networks was also constructed via the Adaboost algorithm. The reason behind the use of the ensemble model of the Adaboost algorithm was to examine if the ensemble model provided superior performance to the single neural network. As indicated by other studies, Adaboost did not have a better performance than a single neural network in this specific application. This may be due to the high noise inherent in the gold data used in this project. Consequently R2 values of gold prediction are also poor for both the single and multiple networks.