摘要
现实世界中存在大量非平衡分类问题,传统K-近邻(K-NN)分类方法采用近邻决策的原则,导致少数类样本在分类过程中难以识别。针对这个问题,提出一种基于聚类的非平衡K-NN分类方法(IKNN_C),该方法通过对已标记的多数类样本进行聚类,压缩多数类样本的规模,从而提高标记样本的平衡性,提高非平衡数据的分类性能。
In real world, there are many imbalanced classification problems, The traditional K-nearest neighbor (K-NN) classification method adopts the principle of nearest neighbor decision rule, and the important minority class samples are always classified wrongly. To solve this problem, this paper presents an imbalanced K-NN classification method based on clustering: (IKNNC). This method can compress the size of majority class size by clustering the majQrity class samples. Then it improves the balance of labeled samples and improves the classification performance of imbalanccd data.
出处
《现代计算机》
2017年第22期6-9,共4页
Modern Computer
关键词
非平衡分类
K-NN算法
聚类
平衡性
Imbalanced Classification
K-NN Algorithm
Clustering
Balance