Physical education is not only an important part of national education but also one of the important means to improve the physical quality of students and citizens. Therefore, the reform of physical education is of great significance to the development of physical education. With the application of data mining technology in the field of physical education, the scale of relevant data increases rapidly. The traditional data analysis methods cannot meet the needs of physical education data analysis. Traditional data analysis methods still have many basic problems to be solved. For example, the professionalism of the structural model and standardization of formal expression are dwarfed by the forefront of the world. There are few real valuable data in the database, and the referentiality is not guaranteed. Therefore, this study puts forward the construction and analysis of discrete system dynamic modeling of physical education teaching mode based on the decision tree algorithm. Through the decision tree algorithm, this study analyzes the data related to physical education and constructs the physical education decision tree system according to the analysis structure. The test results show that the primary influencing factor of physical education teaching is the number of students participating in sports competitions, and the secondary influencing factors are students' liking and teachers' skill level. In addition, teachers' adjustment of physical education teaching contents and methods according to the analysis results of decision tree is conducive to improving students' physical education performance.