计算机工程 ›› 2019, Vol. 45 ›› Issue (2): 167-172.doi: 10.19678/j.issn.1000-3428.0049901

• 人工智能及识别技术 • 上一篇    下一篇

基于参考点的改进k近邻分类算法

梁聪,夏书银,陈子忠   

  1. 重庆邮电大学 计算机科学与技术学院,重庆 400065
  • 收稿日期:2017-12-28 出版日期:2019-02-15 发布日期:2019-02-15
  • 作者简介:梁聪(1991—),男,硕士研究生,主研方向为数据挖掘、大数据处理;夏书银(通信作者),副教授、博士;陈子忠,教授、博士。
  • 基金项目:

    国家重点研发计划(2016QY01W0200,2016YFB1000905);重庆市教委科学技术研究项目(KJ1600426,KJ1600419)。

Improvement k-Nearest Neighbor Classification Algorithm Based on Reference Points

LIANG Cong,XIA Shuyin,CHEN Zizhong   

  1. College of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China
  • Received:2017-12-28 Online:2019-02-15 Published:2019-02-15

摘要:

基本k近邻(kNN)分类算法具有二次方的时间复杂度,且分类效率和精度较低。针对该问题,提出一种改进的参考点kNN分类算法。依据点到样本距离的方差选择参考点,并赋予参考点自适应权重。实验结果表明,与基本kNN算法及kd-tree近邻算法相比,该算法具有较高的分类精度及较低的时间复杂度。

关键词: k近邻, 参考点, 自适应权重, 方差, 分类效率

Abstract:

The basic k-Nearest Neighbor (kNN) classification algorithm has quadratic time complexity,has a low classification efficiency and has a low classification accuracy.Aiming at this problem, an improvement reference points kNN classification algorithm is proposed.The reference point is selected according to the variance of the point-to-sample distance,and the reference point is given an adaptive weight.Experimental results show that compared with the basic kNN algorithm and kd-tree neighbor algorithm,this algorithm has high classification accuracy and has low time complexity.

Key words: k-Nearest Neighbor(kNN), reference points, self-adaptive weight, variance, classification efficiency

中图分类号: