A new algorithm based on fuzzy inference theory is proposed for generating decision rules of pattern classification, which can define a complex nonlinear decision boundary. The effectiveness of the algorithm is demonstrated by a simulation experiment.