Prediction of rock type from physical and mechanical properties by data mining implementations

Rock type characterization is an essential issue in mining and other geosciences. At every stage of mining operations, the rock type is the critical parameter in determining the procedures to be carried out and the equipment to be used. The description of rock types often requires detailed investigations by geologists in the field and laboratory. The experience of the geologists conducting these investigations is also very influential in rock type description. In many cases, this process is time-consuming. With these investigations come extra costs, and, in some cases, relative or inaccurate descriptions can also affect operating costs. This paper shows that it is possible to predict rock type from some physical and mechanical properties of rocks without incurring these costs. The paper's main objective is to present the applicability of data mining algorithms in rock type determination. The physical and mechanical properties of the rocks were evaluated with different data mining algorithms, and the rock types were predicted 95.6% correctly with the model generated with the Support Vector Machine algorithm. Therefore, it is possible to predict rock types by data mining in extensive databases. This method provides both reliable and cost-effective results.