基于平均距离的K-近邻分类改进算法  

Forward a Method Based on the Average Distance K-neighbor Algorithm Classification

作　　者：许燕青[1] 

机构地区：[1]闽南理工学院,石狮362700

出　　处：《电脑编程技巧与维护》2010年第24期41-42,共2页Computer Programming Skills & Maintenance

摘　　要：提出了一种基于平均距离的K-近邻分类改进算法,克服了K-近邻分类算法准确率不高的两个问题:一是各个类别的近邻个数相同时则无法判断测试样本的类别;二是即使某一类别的近邻个数较多,但由于此类别的近邻样本与测试样本的相似度都比较小,则有可能把测试样本错误地判断为此类别。this paper puts forward a method bas ed on average distance K-neighbor algorithm used to overcome the classification,K neighbor-classification algorithm of two reasons not high accuracy of each category：one is the same number of neighbors cannet judge the test specimen category,2 even one kind of other neighbors,but because the number more such other neighbors samples and testing samples are relatively small,similarity is likely to test sample judge this category mistake.

关 键 词：分类 平均距离 K-近邻分类算法 属性值 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]