Abstract: Cognitive physics, from the perspective of natural language, focuses on the thinking process that is from quantity to quality and from data to knowledge as well as the organization of information. In this paper, data mining techniques are explored by using cognitive physics theory and its relevant theory. Firstly, by using the theory of field, the language field is illustrated, which is a method for representing complex knowledge. Secondly, by learning from the information spreading theory, the parameter evolution is discussed, which is an important issue for post-processing of data mining. Finally, by analyzing the information entropy method for decision tree construction, the subjective entropy and the corresponding SID\-3 algorithm are proposed. The advantage of the proposed algorithm is illustrated by using an example.