论文部分内容阅读
针对支持向量数据描述(SVDD)多分类方法中混叠域样本识别精度差的问题,提出了一种提高精度的K近邻隶属度估计算法。首先提取训练样本中的两类混叠样本并在混叠域分别搜寻测试样本的K个近邻,然后通过估计待测样本到K近邻样本中心欧式距离的方法计算样本隶属度,最后通过比较隶属度大小实现样本识别。仿真和UCI数据及模拟电路故障诊断应用验证了算法较传统方法更为有效,尤其适用于不平衡数据的识别。
In order to solve the problem of poor recognition accuracy of aliasing fields in the multi-classification method of Support Vector Data Description (SVDD), an improved K nearest neighbor membership estimation algorithm is proposed. Firstly, two types of aliasing samples in training samples are extracted and K neighbors in the aliasing region are searched separately. Then, the sample membership degree is calculated by estimating the Euclidean distance between the sample to be tested and the K-nearest neighbor sample center. Finally, by comparing the membership degrees Size to achieve sample identification. Simulation and UCI data and analog circuit fault diagnosis applications verify that the algorithm is more effective than traditional methods, especially for the identification of unbalanced data.