Abstract: In this paper，three main factors are extracted to replace seven original factors affecting coal and gas outburst by means of principal component analysis when variance contribution is more than 85%，by which，the input parameters of BP neural network are determined.PCA-BP neural network prediction model is established，which is trained by the study samples from typical coal and gas outburst mines.In order to check feasibility and validity of the PCA-BP model，the instances of a coal mine in Yunnan province are used as predictive samples.PCA-BP model and traditional BP neural network are compared by predictive samples.Simulation results show that the PCA-BP neural network model is superior to traditional BP neural network，and meets the requirement for coal and gas outburst prediction.

Key words: principal component analysis, neural network, coal and gas outburst, forecast

摘要： 通过主成分分析法对煤与瓦斯突出的主要影响因素进行主成分提取，选取贡献率大于85%的3个主成分来代替原来的7个影响因素，以此来确定BP神经网络的输入参数为3个。根据煤与瓦斯突出的类型，建立了煤与瓦斯突出预测的PCA-BP神经网络模型。选用典型突出矿井的煤与瓦斯突出实例作为学习样本，对PCA-BP网络进行训练。以云南某煤矿的煤与瓦斯突出实例作为预测样本，仿真结果表明PCA-BP神经网络模型性能优于传统BP神经网络，能够满足煤与瓦斯突出预测的要求。

关键词: 主成分分析, 神经网络, 煤与瓦斯突出, 预测