提出了一种基于域理论的自适应谐振神经网络算法 FTART,有机结合了自适应谐振理论和域理论的优势 ,以一种独特的方式解决了示例间冲突和分类区域的动态扩展 ,不仅不需要手工设置隐层神经元 ,可以还获得了较快的训练速度和较高的预测精度 .同时还提出了一种可以从训练好的 FTART网络中抽取可理解性好、精度高的符号规则的方法 ,即基于统计的产生测试法 .实验结果表明 ,用该方法抽取的符号规则可以较好地描述FTART的功能 A novel adaptive resonance neural network algorithm based on domain theory FTART is presented. It combines the advantages of adaptive resonance theory and domain theory, and solves the dynamic expansion of conflict and classification region between examples in a unique way. It not only does not require Hidden layer neurons can be manually set up to achieve faster training speed and higher prediction accuracy, meanwhile, a method of extracting symbolic rules with good comprehensibility and accuracy from the trained FTART network is also proposed , Which is based on statistical generation test method.The experimental results show that the symbolic rules extracted by this method can describe the function of FTART