The complexity of the production in coal mine lead to the dynamic characteristics,fuzziness and randomness of coal mine accidents,and impact of coal mine safety risk grade indexes,the complex nonlinear relationship between index and risk level leads to low accuracy of traditional BP neural network evaluation method. This paper puts forward a new method of coal mine safety risk assessment. First,the author created the evaluation index system of coal mine safety risk,and then used analytic hierarchy process( AHP) to determine the index weight,the index for sorting,finally put the RBF neural network as assessment tool and built the coal mine safety risk assessment model based on the AHP of RBF neural network. Through the instance data analysis,the results show that the proposed evaluation model is effective for coal mine safety risk evaluation.