Application of support vector machine in the analysis of the technical state of development in the LGOM mining area

Abstract

The paper presents the results of the analysis of technical wear of buildings located within impact of mining plant in the Legnica - Głogów Copper District (LGOM). The study used method related to neural networks, support vector (Support Vector Machine) in regression approach ε-SVR (Support Vector Regression). The aim of the study was to assess the impact of variables describing the structural protection and renovations on the course modeled phenomenon. The basis for the analysis was created model of technical wear of buildings in the form of a network ε-SVR. In addition to the variables determining the level of structural protection and renovations in the model included variables describing: terrain deformation, mining intensity tremors and the age of the buildings. The choice of model parameters were performed using, as gradientlessness optimization method, genetic algorithm. Based on the established model ε-SVR two types of sensitivity analysis were applied. Assessing the impact of the structural protections have been studying by the analysis of variability of the gradient vector for the modeled hypersurface. The analysis of the impact of renovations on the course modeled process was carried out based on the comparator simulation results of ε-SVR model. The results confirmed the usefulness of the methodology of research and allowed to draw important conclusions on the impact of analyzed factors on the technical wear traditional buildings LGOM.