Research On The Prediction Of Gas Outburst Based On Support Vector Machines

Coal is the main source of China’s resources. Shanxi Province is rich in mineral resources, and is famous with the name "the home of coal". Shanxi occupies a very large proportion of the country’s production of coal in coal reserves. In the safety of coal mine production, gas outburst is the most dangerous. Thus, it is a great theoretical and practical significance that people could make prior accurate warning on gas outburst.The event of gas outburst is a complicated, nonlinear and high dimensional problem, and also synthetically affected by multi factors. It is often difficult to solve with traditional methods. A new way to solve prediction of gas outburst is put forward in this thesis. the latest Statistics Learning Theory of Support Vector Machines(SVM) was used, a prediction method of gas outburst based on the SVM is studied and proposed.This paper presents a support vector machine (SVM) algorithm of data mining to make a prediction and research on gas outburst with the WEKA platform. The SVM is the research focus in recent years, which solves the practical problems, such as small samples, nonlinear, high dimension and the local minimum value, and also has a good classification and recognition effects with the basis of statistical theory. Because of its excellent learning and generalization performance, SVM attracts great concern and research and has been widely applied to various fields. WEKA open source software is developed by Waikato University using JAVA language, which contains a variety of mining algorithms to facilitate people learn and research.The paper implements the combination of quantitative and qualitative through the introduction of the SVM to analyze the prediction of gas outburst. That is to say, the monitored data can be used to detect the gas outburst, and to reflect the classification in real world. At first the data monitored by surveillance is pre-processed. Due to the uncertainty and vagueness of the influencing factors of gas outburst, gray relational theory is adapted to sort the outstanding influencing factors in the order of influencing degree from high to low. We choose the required influencing factors as the feature vectors. In order to eliminate the difference between the various influencing factors and normalize the data, the required data table is obtained. We successfully integrate the LISVM software into the WEKA software, and implement the SVM algorithm. Besides, WEKA is used to predict the gas outburst. The existing case is made for modeling in the WEKA platform using the cross method. The unknown gas monitoring data could be predict and classified through modifying the parameters and several experiments.Through the application, support vector machine algorithm is reasonable and effectiveness for the gas outburst in the platform of WEKA, which could make a correct prediction to the gas outburst. Thus, the mining stuff could response in time to avoid the great loss of personnel and property. The method provides reference value to further explore the practical and effective prediction of coal and gas outburst.