摘要
k-最临近(k-NN)分类方法在计算两训练样本的相异度时给每一属性加相同的权,这样会造成分类的准确性下降,尤其当存在很多无关属性时,甚至会造成混乱。针对这一弱点该文提出了一种用每一属性的信息增益作为该属性的权来计算训练样本间的相异度的数学模型,并将这一模型应用于k-最临近分类方法,改善了该方法的分类质量。
In k-NN method same factor is assigned to each attribute when computing the dissimilarity of two different training samples,as a result,the accuracy of classification gets worse and even confusion will come into being especially when there are many irrelevant attributes.In order to eliminate that weakness in this paper a mathematical model assigning information gain as factor to each attribute when computing the dissimilarity of two different training samples is put forward and applied to k-NN method,consequently the classification quality of this method is improved.
出处
《计算机工程与应用》
CSCD
北大核心
2003年第35期183-185,共3页
Computer Engineering and Applications
