Pattern Recognition And Integrated Decision Of Coal And Gas Outburst Disaster Information

China is a major coal producing country, unfortunately, the coal and gas outburstdisasters occur frequently. This constraints on lives and coal mine safety. This articleuses pattern recognition to process outburst information, and to construct outbursthazard prediction models, which provides decision-making services to the users.From the perspective of system engineering, establish the coal and gas outburstspace-time forecasting system. In time domain, carry on early comprehensive analysisand run-time forecast; in the space domain, carry on regional prediction, localmonitoring and point prediction. According to principle of contact and development,process the outburst information on the overall level.For different stages of prediction, combined with the existing mechanism ofoutburst prominent, select the appropriate set of prediction indexes. Forward floatingsearch, optimization algorithms, rough set theory are used to get attribute reduction.During the prediction model building phase, I take into account the restrains of NOFREE LUNCH theory, fully consider various possible situations, and search theliterature, so one conclusion obtained, there is no absolute good and usefull algorithms,also do not have the best and most comprehensive feature sets. Even the correctalgorithm selected, we may not be able to fully solve complicated practical problems.Therefore, it is necessary to select the simple and effective one of available methods,even establish the multiple algorithms fusion model. We select neural network, SVMetal, what’s more, I improve the feedforward neural network classifier by using chaosbee colony algorithm(CBC-MLP). However, coal and gas outburst prediction is acomplicated system, it’s difficult to use single algorithm to solve all the problems. Sobase on established integrated model, more useful prediction results will be gotten.The last part of the paper is the engineering realization of a series of algorithms,classifier models. Pattern recognition module and decision support modules aredesigned. Combined with actual situation at the scene, use VC++6.0as developmenttools to design CGOCIDS. Call monitoring data, integrate algorithms programmed byMatlab, to complete the identification process and establish prediction model.