摘要
为了提高近邻法的分类性能,提出了核的最近邻算法。通过mercer核,将样本映射到高维特征空间,再用近邻法分类。核映射改善了样本的空间分布,突显了样本的类别特征,从而提高了分类的性能。给出了核近邻算法的判决过程。对于人工数据和入侵检测数据的仿真显示,核近邻分类方法的分类性能优于传统的最近邻分类法。
In order to improve the classification ability of the nearest neighbor rule, a kernel nearest neighbor rule is presented. Using the mercer kernel, the samples are mapped to high dimensional feature space, and are classified there. By kernel mapping, the distribution of samples is improved, and the features of samples are stand out. Thus the performance of the kernel nearest neighbor rule is improved. The process of discrimination using kernel nearest neighbor rule is given. The simulation results using both sysnthesized data and intrusion detection data show that the KNN rule has better performance than the NN rule.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2007年第3期470-471,共2页
Systems Engineering and Electronics
基金
"十五"军事通迅预研基金资助课题(41001040303)
关键词
模式识别
核方法
近邻法
入侵检测
pattern recognition
kernel method
nearest neighbor（NN） rule
instrusion detection
