Use of machine learning techniques to analyse the risk associated with mine sludge deposits

Sludge deposits resulting from mine extraction activities and accumulating in the proximity of production centres have an important potential impact on their surroundings. This potential impact needs to be evaluated by quantifying the risk of an accident on the basis of a joint study of factors affecting the probability of occurrence, environmental, populational and infrastructural vulnerability factors and intrinsic and extrinsic risk factors. The problem is non‐linear, and this fact, combined with the high number of risk conditioning variables, justifies using machine learning techniques to estimate risk. A comparison of results for supervised versus non‐supervised learning techniques confirms that the former adapts better to the problem than the latter.