Abstract: Aiming at the problems of the traditional coal and gas outburst early-warning system, such as insufficient self-learning ability, low information fusion, poor intelligence level and low early warning accuracy, based on correlation analysis and evidence theory algorithm, an independent dynamic optimization method of prominent early warning indicators and a multi indicator fusion decision-making model are established, and a prominent intelligent early warning system is developed. The construction mode of hazard identification affairs set is put forward; the asymmetric binary processing method for early warning indexis established; the construction method of association rules between early warning indicators and outburst itemsis put forward; a function reflecting the size of the false alarm rate is defined; the optimal method of warning index based on association rule analysis is formed. A evidence theory identification framework of coal and gas outburst early warning fusion analysis is put forward. The basic probability distribution function of evidence body is constructed; the synthesis method of evidence is given; a function reflecting the danger probability is constructed; a multi-index fusion decision-making method for coal and gas outburst early warning has been formed. On this basis, the intelligent early warning system of coal and gas outburst based on service architecture model is developed, which has the functions of online monitoring, intelligent analysis, integrated early warning, online release and remote query. The field application results show that the accuracy rate is 91.53%, and no missing alarm. It provides a good technical support for the prevention and control of coal and gas outburst.