Abstract: In order to fully learn the experience of coal and gas outburst historical accidents and give full play to the value of expert opinions in accident investigation reports, an CBR outburst early warning model based on case-based reasoning was proposed, and an outburst early warning system was built based on historical case database. With the multi-category index data of outburst as input, the K-nearest neighbor algorithm was used to calculate the local similarity between the current instance and the past instance. At the same time, in order to further improve the accuracy of case retrieval, the grey wolf optimization (GWO) algorithm was used to optimize the feature weight of each index, and the global similarity was calculated. By conducting similarity matching between the current instance and the past instance, the outburst risk was early warned, and the prevention and control decision scheme of outburst was put forward. Using the outburst accident in Hemei No. 6 Mine in Henan Province to verify, the results show that the established early warning system can realize effective early warning and decision-making of outburst danger.