Research On Prediction Method Of Coal And Gas Outburst Based On The Entropy And SVM

Coal and gas outburst,which is one of the major accident of coal mine,can cause hugeeconomic losses,casualties and other undesirable social influence. However, Coal and gasoutburst is a complicated nonlinear problem, and is influence of various factors, such assurrounding rock conditions, coal properties and geological structures. The traditionalforecasting method have certain limitation, low accuracy and hard to application. Therefore,how to make an effectively forecast analysis becomes the concentration of coal mine safetyworkers.This paper firstly summarized the entropy theory, the calculation of entropy weight andfiltered the related historical data of coal and gas in certain coal mine. Secondly, expoundedthe principles and basic theory of Support Vector Machine (SVM), described respectively themodel of linear classification problem, approximate linear classification problem andnonlinear classification problem, and thus provided the basis for the coal and gas outburstforecast model construction. Thirdly, discussed the basic rules and influence factors of coaland gas outburst, listed the current coal and gas outburst prediction index and use entropyweight to process the coal and gas outburst data of certain coal in Shanxi province. Fourthly,used malab to achieve the coal and gas outburst forecast model and predict the optimizationdata. The result showed that this method had a higher prediction accuracy and betterpracticability. At last, used GUI graphical interface design function of matlab to show the coaland gas outburst forecast model.Based on the research results, some conclusions could be drawn as follows: it’s feasibleto apply SVM into coal and gas outburst forecast. The prediction method of coal and gasoutburst based on the entropy and SVM in this paper has high accuracy and good application.The predict result showed there was certain threat of coal and gas outburst in this area. Aftermodifying the corresponding parameters and data sets, this model could be applied to other classification problem.