摘要
初始聚类中心的随机选择,根据主观经验确定类簇数等问题时常伴随着原始K-means算法。为了攻克以上问题,改进算法采用峰值法以及融合了K近邻算法的密度峰值算法逐一调整。通过在UCI数据集上测试及与原始K-means算法、最大最小距离距离算法在准确率、稳定性和处理数据速率方面的比较,其中最为突出的是,改进算法的准确率达到了96%以上。
The random selection of the initial clustering centers,and the determination of the number of clustering based on subjec⁃tive experience often accompany the original K-means algorithm.In order to overcome the problems,the algorithm used Peak meth⁃od and the fusion of the density peak algorithm and K-nearest neighbor algorithm to adjusted K-means algorithm.The most promi⁃nent of these is that the accuracy of the improved algorithm has reached more than 96%through testing on the UCI dataset and comparing with the original K-meaning algorithm,the maximum and minimum distance algorithm in terms of accuracy,stability and processing data rate.
作者
王炜唯
周云才
WANG Wei-wei;ZHOU Yun-cai(School of Computer Science,Yangtze University,Jingzhou 434020,China)
出处
《电脑知识与技术》
2021年第8期182-184,共3页
Computer Knowledge and Technology
基金
2019年中央引导地方科技发展中的油田数据智能分析研究资金项目(2019ZYYD016)。