A modeling method based on matrix singular value decomposition(SVD) for model structure identification and parameter estimation is proposed for the complex, uncertain and non-uniformly sampled data nonlinear systems. Firstly, the matrix singular value decomposition method is used to analyze the relationship among the local model and the accumulation contribution rates, and the rule number of the fuzzy model is determined. Thus, the structure optimization of the model is realized. Then, in order to overcome the error accumulation and transfer of recursive least squares, the conclusion parameters of the recursive least square estimation algorithm with singular value decomposition is adopted. Finally, a simulation example is given to illustrate the effectiveness of the proposed algorithm.