Mine workers threshold shift estimation via optimization algorithms for deep recurrent neural networks

Abstract

Abstract The application of deep learning techniques to estimate mine workers threshold shift is investigated. The Recurrent neural networks is used to estimate the hearing threshold shift of mining employees. A critical analysis of the different optimization methods is performed. The Adaptive Sub-gradient Method optimization is preferred over the other methods due to its fast rate of convergence. The recurrent neural network predicts the threshold shift with an accuracy of 95%. The obtained results can used in the development of an early intervention and monitoring system for the mines. The future performance of the model can be improved by including more inputs to the system.