摘要
针对多故障样本一次性映射之后分类不理想,研究了多级层次式支持向量机,应用UCI数据仿真,结果表明该方法缩短了训练时间、提升了测试准确率、改善了样本的可分类性。由于误差积累,在此基础上,提出了层间嵌入式多级支持向量机,采用Abalone数据仿真,结果表明分类精度有所提高并减少了分类步骤。结合ReliefF算法,对制导设备的故障特征参数进行了选择。
Multi-grades SVM is studied. This method uses UCI data emulation mode, and the result shows that this method shortens training time, improves the test veracity and promotes the classification nature of the sample. Because of the error accumulation, a multi-grades support vector machine is put forward and Abalone data emulation mode is adopted. The result indicates that the classification precision is raised and the classification step is reduced to some extent. By using ReliefF＇ s algorithm, the selection is conducted to fault characteristic parameter of the missile guidance component.
出处
《军械工程学院学报》
2010年第4期9-12,共4页
Journal of Ordnance Engineering College
关键词
故障
多类分类
支持向量机
特征选择
设备
fault
classify
support vector machines
feature selections, equipment
