Gom-Svm Predictor For Land Subsidence at Finished Underground Mining

Abstract

Underground mining is one of the causes of land subsidence. The process of the land subsidence caused by underground mining is complicated and systematicness. Accurate prediction the land subsidence has important practical or immediate significance to avoid the harm of land subsidence. Grey system theory was applied extensively and had gained a series of achievements in land subsidence prediction, but our preliminary study showed that the general GM(1,1) model was inadequate to handle prediction as its only adapt to the data with exponential law. The advantages and disadvantages of general GM (1, 1) model and support vector machine (SVM) are analyzed respectively. A new land subsidence forecasting model based on GOM (1, 1) and SVM model was put forward. The new model develops the advantages of accumulation generation in the grey forecasting method, weakens the effect of stochastic disturbing factors in original sequence, strengthens the regularity of data, and avoids the theoretical defects existing in the grey forecasting model. The advantage of support vector machine which can fit nonlinearity time series data efficiently was also used. The example shows that the prediction accuracy has been improved quite a lot in comparison with general grey model.