摘要
通过肘方法确定类别数,采取平方差半径法选择聚类种子中心,优化聚类中心的重新选择,采用熵权法对数据对象的属性赋权修正对象间的欧式距离,计算属性间的作用差异。结果表明,在类别数不变,添加异常数据后,对于维度低、类别间差异大的小样本数据,改进算法在执行效率几乎等同的情况下比原算法精确、稳定。
Firstly,the number of categories was determined by the elbow method,and then the square difference radius method was used to select the cluster seed center,and the re-selection of the cluster center was optimized.The entropy weight method was used to weight the attributes of the data objects to correct the Euclidean distance between the categories and calculate differences in the role of features.The results show that after the number of categories remains unchanged and abnormal data is added,for small sample data with low dimensions and large differences between types,the improved algorithm is more accurate and stable than the original algorithm with almost the same execution efficiency.
作者
刘畅
肖斌
蒋铁军
苏凯
何鹏翔
王成宇
LIU Chang;XIAO Bin;JIANG Tiejun;SU Kai;HE Pengxiang;WANG Chengyu(Department of Management Engineering and Equipment Economics, Naval University of Engineering, Wuhan 430033, China;Armed Police Second Mobile Corps, Fuzhou 350200, China)
出处
《兵器装备工程学报》
CSCD
北大核心
2021年第S01期266-270,共5页
Journal of Ordnance Equipment Engineering
基金
国家自然科学基金项目(61802425)
国家社会科学基金军事学项目(16GJ003-105)。
关键词
信息熵
K-MEANS
小样本
欧式距离
聚类中心
肘方法
误差平方和
聚类精确度
information entropy
K-means
small sample
euclidean distance
cluster center
elbow method
sum of squares of error
clustering accuracy
