Selection of parameters of the support vector machine method to the problem of subsidence modelling due to drainage

Abstract

This article presents the problem of modelling drainage subsidence that accompanies the mining of solid minerals. Rock mass drainage causes a change in pressure in the aquifer, and thereby initiates the compaction process. On the surface we can observe the effect in the form of a wide drainage basin, which adds to the direct impact of mining operations. The article presents the research stage associated with the use of artificial intelligence in forecasting the indirect impacts of (drainage) in mining areas. This article also outlines the Support Vector Machine (SVM) method and its use based on the example of underground coal mining. For the purpose of calculations, the data from altitude surveying conducted on the terrain surface, and information from the network piezometric boreholes installed in subsequent aquifers were used. Used in the analysis was ε-SVM method for regression tasks with the use of radial basis function. The calculations were performed with an integrated software package for support vector regression (LIBSVM) and the obtained results were presented. The process of selection of parameters in different variants, and obtained discrepancies in the process of research and testing were described. Cross-Validation and generalization of the knowledge processes necessary for future forecasting the process of drainage subsidence were characterized. The summary includes opportunities for further research as well as analysis using artificial intelligence.