Abstract: The paper introduced principle of BP artificial neural network and Bayesian regularization(algorithm,) discussed establishment of model of BP artificial neural network utilizing Bayesian regularization,established the best network structure of prediction model of coal and gas outburst which had only one hidden layer with only one neural node through experiment of changing quantity of neural node in(hidden) layer of the network.The result of using prediction indexes of coal and gas outburst to train and check the network showed that the danger degree of coal and gas outburst of prediction by the netwoprk was according with the actual status completely.The result of analyzing inputted prediction indexes of coal and gas outburst of input layer of the network and influence weight of outputted prediction result of output layer showed that the geological structure type of coal seam had the largest effect to coal and gas outburst.Above research results had some reference value to research of prediction and prevention of coal and gas outburst and improvement of its prediction accuracy.