在对大型数据库聚类中的伪装危险数据进行识别时,由于具有的模糊性,使得数据的特征之间存在关联性较高。传统的伪装危险数据识别过程,由于在高关联性下识别伪装危险数据,容易产生多中心问题,具有较高的误警率,提出采用自组织特征映射的大型数据库聚类中的伪装危险数据识别方法,确定可形成映射的自组织特征映射模型,用于描述伪装危险数据的特征。上述模型可使神经元的权系数形态间接模仿输入的信号模式。将训练数据作为一个输入向量输入自组织特征映射网络中,从显示邻域保留的网络中形成映射图,对参数进行分析,在输出图中获取伪装危险数据,完成大型数据库聚类中伪装危险数据的识别。仿真结果表明,所提方法具有很高的识别精度。
A recognition method of camouflage risk data in large database clustering is presented based on the self -organizing feature map, a mapping model of self-organizing feature established to describe camouflage dangerous data characteristics, the model can make the weight coefficient of neurons in the form of indirect imitate input signal mode. The training data is the input vector of the self-organizing feature map network, and the map is formed from the display neighborhood preserving network, its parameters are analyzed, the camouflage risk data in the output graph are obtained, and the recognition of camouflage risk data in large database clustering is completed. The simulation results show that the proposed method has high recognition accuracy.