An effective and general model — projection pursuit clustering model is proposed in this paper. It can solve the incompatible problems of indexes of safety evaluation，which are uncertain，fuzzy and complex. At the same time, it improves the recognition rate of the coal mine safety evaluation model. It uses genetic algorithms to find optimal solutions for the model, the information of which can be used to research the influence of all the indexes on the evaluation of coal mine safety. The paper clusters the safety level of each coal mine according to each projection value, which provides the objective basis for decision-making of the evaluation of coal mine safety. This paper combining with the example demonstrates operation process. It also proves the method is scientific and feasible, which is of significance of coal mine safety evaluation.