摘要
运用结构风险最小化原理和聚类原理,将支持向量机中有监督的分类算法与统计中无监督的聚类算法有机地结合起来,对线性可分与线性不可分两种情况分别建立了无监督的分类模型.模型的求解转化为一个二次规划问题,同时此模型也适合于多分类情况.在应用到心脏病的医疗诊断中,准确率为88.5%,较以前的方法有了明显的提高.
A kind of unsupervised classification algorithm based on combining support vector classification algorithm with statistical clustering algorithm by following structural risk minimization principle and clustering principle is proposed. The solution of the model is transformed into a quadratic programming problem. The model is fit for multi-classification problem too. It can be applied in medical inspection of heart diseases and the veracity is 88.5%, which is better than those of other methods.
出处
《烟台大学学报（自然科学与工程版）》
CAS
2004年第1期9-13,共5页
Journal of Yantai University(Natural Science and Engineering Edition)
关键词
SVM
分类算法
支持向量机
聚类算法
期望风险
结构风险
support vector machine
classification algorithm
clustering algorithm
expectation risk
structural risk minimization
