Abstract: Based on the binary tree structure and fuzzy logic theory, a fuzzy-tree model applied to complex system modeling is proposed in this paper. Linear models and fuzzy sets are arranged in a tree structure. A hybrid-training algorithm is used to update linear model coefficients and membership functions of the fuzzy sets. Compared with some other modeling methods, such as ANFIS, the proposed model is of less computation, higher accuracy, especially for high dimension data modeling. Simulation results demonstrate the advantages of this approach.