Abstract: In order to improve the accuracy of coal and gas outburst prediction, we extract the core index based on the rough sets approach to the original outburst data of outstanding multi-factors and indexes. The core multi-factors and indexes are displayed with the set pair theory by the integrated use of a specific connection degree mathematical model, and the indexe weight is determined by the entropy method. Combined with the clustering analysis, we identify the category of the system under test sample by calculating the difference and the distance between the contact agree of forecast system according to the difference and the pattern recognition of "choosing close principle". Take Yanshitai Mine as an example, the model predicted the test sample, and the results conform to the actual situation.