Data mining is one of the techniques of taking information of unordinary and unknown potential value out of large-scale data. The main idea of rough set method of data mining is to divide knowledge with undistinguishable relationship, to describe the concept with the upper and lower approximation, and to reach the rules of decision-making or classification through the reduction of knowledge. By means of examples, this paper introduces the essential principles and methods of rough set theory, and applies the mining classification rules to the differentiation of symptoms and signs of traditional Chinese medicine (TCM) for cancer patients. The application of rough set theory in TCM helps reveal the intrinsic relationship of knowledge in TCM diagnosis and the thinking stringency of TCM, providing theoretical security for the differentiation of cancer patients treated by TCM.