提出了一种基于神经网络的自组织模糊系统，它能够根据输入输出数据灵活地划分模糊集合．由于采用模糊聚类方法和梯度下降法分两步对该系统进行训练，其收敛速度要比传统的ＢＰ算法快得多．仿真结果表明该系统结构简单，学习速度快，规则数少，模型精度高 A self-organizing fuzzy system based on neural network is proposed, which can flexibly divide fuzzy sets according to input and output data. Since the system is trained in two steps using fuzzy clustering method and gradient descent method, its convergence speed is much faster than the traditional BP algorithm. The simulation results show that the system has the advantages of simple structure, fast learning speed, few rules and high precision