In this paper, we present one entropy partition method and its ideal requirement for partition result, then apply this method in TCM (Traditional Chinese Medicine) data without understanding any expectation of the objectives. We propose the conception of N-class correlation in order to solve the computational problem successfully. The idea of RFS (relatives and friends set) and general filtering are also helpful to further data mining research. The result achieves a good explanation in TCM theory, thus validates the syndromes existence in western medicine. It also shows that this method is rather available especially in dealing with the complex, nonlinear, large-scale data