Noise level policy advising system for mine workers

A noise level policy advising system to be used by mine administrators in assigning tasks to new employees at the mine is proposed. The presented novel system uses machine learning techniques which includes clustering and classification of new employee. The mine workers are clustered using K-means to determine their properties. By comparatively using Logistic regression, support vector machines, decision trees and random forests classification techniques, the mine workers are classified. Depending on the classification, which is based on the mine workers baseline and future threshold shift, recommendations to suitable mining tasks are made. The decision tree is the best performing model with the highest accuracy. It has an average testing accuracy of 91.25% and average training accuracy of 99.79%. However, logistic regression provides the best generalisation results on the testing set. Future work would include development of a friendly Graphical User Interface to facilitate easy use of the system.