Abstract: In order to solve the problem of relevance and fuzziness in coal mine workers’ unsafe behavior evaluation model better, this paper constructs the evaluation index system of coal mine workers’ unsafe behavior from five aspects including management factors, working environment, mechanical equipment, staff quality and team and group factors. We use analytic network process to study the interaction among indicators, and determine the weight of indicators. The analytic network process and matter-element extension model are combined to evaluate the factors that affect the unsafe behavior of coal mine workers. Applying this evaluation model to assess the unsafe behavior of workers in L Coal Mine, the results show that the grade of unsafe behavior of employee in L Coal Mine is Grade Ⅱ, the employee unsafe behavior status is good, and the result of assessment is consistent with the actual Coal Mine, which indicates that the model has a certain feasibility and scientific.