摘要
聚类分析是机器学习和模式识别领域的一个重要问题,聚类算法常用于解决这类问题。针对传统聚类算法运算量大、不适应任意分布数据聚类的不足,提出了一种基于数据包含度的自动聚类算法。该算法引入数据包含度的概念,能够自动确定聚类个数和聚类中心,并进一步采用跟随策略实现聚类。多组数据的实验验证了自动聚类算法的有效性。对不同分布的数据进行了自动聚类算法与K-means聚类算法的聚类结果比较,实验结果表明自动聚类算法具有很好的聚类性能。
Cluster analysis is an important issue for machine learning and pattern recognition. Clustering algorithm is usually used in solving these problems. A novel automatic clustering algorithm is developed based on data contained ratio. In automatic clustering algorithm which is presented in this paper,the concept of data contained ratio is proposed,the cluster number can be determined automatically based on the data contained ratio,and the relative cluster centers are found similarly Several groups data are used to testify and demonstrate the validity and effectiveness of the cluster algorithm. In addition,the comparison between the traditional K-means cluster algorithm and automatic cluster algorithm is processed. The results demonstrate that the automatic cluster algorithm has high performance in clustering random distribution data set.
作者
马云红
王成汗
江腾蛟
张堃
Ma Yunhong Wang Chenghan Jiang Tengjiao Zhang Kun(School of Electronics and Information, Northwestern Polytechnic University, Xi'an 710072, China)
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2016年第5期863-866,共4页
Journal of Northwestern Polytechnical University
基金
西北工业大学研究生创意创新种子基金(G2015KY0407)
国家自然科学基金青年基金项目(61401363)资助
关键词
聚类算法
数据包含度
数据局部密度
clustering algorithm
data contained ratio
data local density