Abstract: To improve the level of safety integration supervision in county mining areas, and to solve the problem of weak macro-control ability of regional-level coal mine safety at the national regulatory level, using field surveys and expert consultation methods, based on the six levels of “human-environment-coal mine-information-supervision-rescue”, we established a county mining area safety risk evaluation index system, established a evaluation model of coal mines in a region based on ANP and probabilistic neural network, and conducted an empirical study. Research results show that the index system constructed is reasonable and effective and the evaluation model has higher accuracy, it can provide more accurate early warning evaluation of county mining area safety.