Application Resrarch On The Coal And Gas Outburst Prediction Based On Gray Correlation Analysis And PSO-SVM

Coal and gas outburst prediction is a multidimensional,complex and nonlinear prediction system.With the rapid development of data mining technology and machine learning,some intelligent machine learning methods can be used to effectively solve this type of problems.They are introduced into the prediction of coal and gas outburst with high adaptability.This paper puts forward seven factors that have a great influence on the coal and gas outburst by summarizing and analyzing the mechanism of coal and gas outburst.On this foundation,combined with the actual situation,the parameter system for predicting coal and gas outburst is established.However,if so many parameters are used as input to the prediction model,the model will become complicated due to the large dimension of input parameters,which will affect the calculation efficiency and reduce the generalization ability of the model.The grey correlation analysis method in grey system theory is introduced to analyze the degree of correlation between each parameter and the research problem.Through the calculation of the degree of correlation,the main controlling factors that can reflect the research problem strongly are selected according to the calculation results.It can achieve the purpose of dimension reduction.The main control factors are selected as input to predict coal and gas outburst.The sample data collected in mine are divided into training sample set and test sample set.The sample set is used to train the model and the test set is used to verify the prediction accuracy of the model.Firstly,support vector machine(SVM)prediction model is used for the training and prediction.The determination coefficient of the predicted results is 0.90935.At the same time,BP neural network model is used for training and prediction.Through the comparison and analysis of the predicted results of two type of models,it shows that support vector machine has a higher prediction accuracy.It indicates that SVM has higher adaptability than BP neural network in the prediction of small sample data.The combination of penalty factor C and kernel parameter g in SVM kernel function will affect the prediction accuracy of the model.Since the traditional methods are easy to fall into the dilemma of local optimal solution when looking for the combination of C and g.In this paper,particle swarm optimization(PSO)is introduced to optimize the two parameters of SVM.Particle swarm optimization(PSO)can minimize the occurrence of local optimal solutions by optimizing parameters from a global perspective.The PSO-SVM prediction model is used to train and predict for coal and gas outburst.Meanwhile,the model optimized by genetic algorithm is used for training and prediction.The determination coefficient of prediction results of PSOSVM model is 0.93654,which is high than GA-SVM model.The results show that both models have good effects.However,compared with genetic algorithm,particle swarm optimization algorithm has the advantages of simple structure,fast convergence,high computational efficiency and high generalization ability.In addition,the stability of pso-svm model is tested,and the test results show that the model has high stability.The predicted results are verified,and the results show that the predicted results of the PSO-SVM model are consistent with the actual situation.To sum up,it shows that it has higher application value in coal and gas outburst prediction.