Prediction Of Coal And Gas Outburst Under The Strategy Of Unbalanced Data

With the increase of coal demand,the power disaster accidents of coal and gas outburst occur frequently and become more and more serious.Therefore,it is necessary to study the risk prediction of coal and gas outburst in order to realize its fast and accurate prediction.However,in real life,the events of coal and gas outburst are far less than those of no outburst,and the outburst intensity is also different,so the collected data samples are unbalanced data,and the imbalance of data sets directly affects the accuracy of classification and prediction.Therefore,this paper mainly deals with the problems of coal and gas outburst from two aspects of data and algorithm under the strategy of unbalanced data response The IFOA-ELM prediction model based on GREA-PCA data processing was proposed.Firstly,we adopted undersampling strategy,oversampling strategy or change evaluation index strategy for unbalanced data.Then,we mainly used grey relational entropy(GREA),which is a combination of grey relation analysis and entropy weight method,to extract the main influencing factors,and use principal component analysis(PCA)to reduce the dimension.In the algorithm,we mainly improve the initial position dimension,individual search step length of fruit fly optimization algorithm and introduce the group fitness variance in iterative updating to form IFOA algorithm and optimize the weight and threshold of limit learning machine,and build IFOA-ELM prediction model.Finally,the simulation experiment was carried out on the python platform by using the UCI data set fertility data set.It is found that the prediction accuracy of oversampling strategy and changing the evaluation index strategy is higher than that of undersampling strategy.Under the three strategies,the data processed by GREA-PCA is better than that without GREA-PCA.Compared with IFOA-ELM model,FOA-ELM model and ELM model,IFOA-ELM model is better and more accurate than ELM and FOA-ELM model.Through the analysis of the data samples of coal and gas outburst in Pingdingshan No.8coal mine,nine influencing factors of coal and gas outburst were selected synthetically.Under the strategy of over sampling and changing the evaluation index,the data was input into IFOA-ELM model based on GREA-PCA data processing,and the comparative experiment was carried out.The results show that the data processing of GREA-PCA can effectively improve the prediction effect of the model,whether it is oversampling strategy or changing the evaluation index strategy.Compared with FOA-ELM model and ELM model,the prediction results of IFOA-ELM model are consistent with the actual results.Therefore,it can be shown that the IFOA-ELM model based on GREA-PCA data processing has high accuracy in the prediction ofcoal and Gas Outburst under the strategy of unbalanced data processing,which provides a new method for effectively realizing the prediction of coal and gas outburst.There are 22 figures,33 tables and 73 references in this thesis.