In the practical applications of genetic algorithm(GA), a lot of problems can be described as tree structures. In fuzzy modelling, the structure identification of a fuzzy model is to partition the input domain, which is in essence a process of complicated nonlinear optimization. The partition of the input domain can be expressed as a binary tree by a fuzzy tree model. A matrix coding approach to representing a binary tree is presented, which is simple and suitable for the genetic operations in GA. The partition of the input domain is optimized by GA with matrix coding for tree-structured individuals and theresulting suboptimal fuzzy tree model is obtained, which has higher precision and lower complexity of model. A simulation example is given to validate the proposed method.