Human resources management (HRM) is very important for the enterprises' existence and development, but how to measure the HRM performance is a major issue that troubled the enterprises. This paper overcomes the shortcoming of tradition linear HRM evaluation method, proposes a evaluation method which unifies the BP neural network algorithm and the multistage dynamic fuzzy judgment, takes the multistage dynamic fuzzy judgment as the sampling foundation, uses the BP neural network principle to establish evaluation model. This method not only can exert the unique advantages of BP neural network, but also overcome the difficulty of seeking the high grade training sample data. The HRM performance evaluation of 10 enterprises, indicates that the method to evaluate the HRM is stable and reliable.