By means of discovering crosscutting concerns from legacy systems, aspect mining intends to help migrate the systems to an aspectoriented design. An improved method based on maximum tree method for aspect mining is presented. The method uses aspect ideas to capture the runtime methodcall information by mining crosscutting concerns from dynamic behaviors, and then constructs a methodcall relationship data matrix. Based on fuzzy similarity relation theory, by introducing the similarity, an object similarity matrix is constructed, and the maximum tree method is used to identify the crosscutting concerns in the system. The method can provide a basis for system’s software reconstruction and reusability. An experiment is conducted to verify the validity of the method. Compared with the existing typical mining methods, the method shows the virtue of clear implementation and high efficiency.