一种粗糙模糊神经网络识别器及其应用

A Fuzzy Neural Network Classifier Based on Rough Set and Its Application

结合粗糙集和模糊神经网络提出了一种粗糙模糊神经网络识别器的模型.该模型根据粗糙集理论对训练样本进行建立决策表、离散决策表、约简决策表、提取分类规则等推理过程设计.粗糙模糊神经网络识别器的输入层、两个隐含层、输出层的神经元个数分别根据决策表的约简结果、离散结果和分类规则、决策属性决定.将该识别器用于车牌字符识别,实验表明:该方法比粗糙集规则匹配识别方法识别率提高了18%,比BP神经网络识别方法识别率提高了2.7%.

A model of rough fuzzy neural network classifier was presented by combining rough set and fuzzy neural network. The model was designed according to the reasoning process of the rough set theory, including building up, dispersing and reducing the decision table of training samples. In the rough fuzzy neural network classifier, the number of the neurons of the input layer, two hiding layers and the output layer were determined by the reduction results, the dispersal results, classification rules and the decision attribute of the decision table, respectively. The application of the classifier to the car‘ s plate characters recognition shows that it can increase the recognition rate of 18% compared with the method based on rough set rule matching, and 2. 7% compared with the method based on BP neural network.