Application of predictive data mining to create mine plan flexibility in the face of geological uncertainty

Abstract

Abstract Geological uncertainty represents an inherent threat for all mining projects. Mining operations utilise resource block models as a primary source of data in planning and in decision making. However, such operational decisions are not free from risk and uncertainty. For the majority of iron ore mines, as an example, uncertainties such as clay pods and variability in grades and tonnages can have dramatic impacts on projects’ viability. However, a paradigm shift on how uncertainty is treated and a willingness to invest in areas that create operational flexibility can mitigate potential losses. Data analytics is touted as one of the major disruptions in the 21st century and operations that properly utilise data can create real opportunities in the face of an uncertain future. Since organisations have abundant definite geological data, a combination of data mining and real options can provide a competitive advantage. In the present study, predictive data mining algorithms were applied to a real case mine operation to predict the probability of encountering problematic ore in a mining schedule. The data mining model outputs were used to generate possible real options that the operations could exercise to deal with clay uncertainty. The most suitable data mining algorithm chosen for this task was the classification tree, which predicted the occurrence of clay with 78.6% precision. Poisson distribution and Monte Carlo simulations were applied to analyse various real options. The research revealed that operations could minimise unscheduled losses in the processing plant and could increase a project's present value by between 12% and 21% if the predictive data mining algorithm was applied to create real options.