摘要
针对DBSCAN算法的不足,提出了一种基于DBSCAN的自适应聚类算法.通过引入对象密度迅速地找到数据集中的核心样本,并从核心样本出发进行统计学分析得到Eps与MinPts之间的函数关系及相关的Eps与MinPts参数值,并利用所获参数值进行自适应的聚类;采用若干个仿真和真实数据集进行实验,评估该算法的有效性和可靠性.实验结果表明,该算法对密度不均匀数值型数据集和符号型数据集均有较好表现.
Aiming at the shortcomings of DBSCAN,an adaptive clustering algorithm based on DBSCAN was proposed.This algorithm introduces a concept of object density,which can quickly find the core sample in the data set.Starting from these core samples,the functional relationship between EPS and Minpts was not only gotten,but also the values of EPS and Minpts are obtained.The adaptive clustering was carried out by using the obtained parameters.Finally,in order to evaluate the effectiveness and reliability of the algorithm,several simulation and real data sets are applied for t experiments.The experimental results show that the algorithm performs well on non-uniform and symbolic data sets.
作者
陈小辉
奚庆港
CHEN Xiao-hui;XI Qing-gang(College of Computer Science and Technology,Huaiyin Normal University,Huaian Jiangsu 223300,China)
出处
《淮阴师范学院学报（自然科学版）》
CAS
2021年第3期228-234,共7页
Journal of Huaiyin Teachers College;Natural Science Edition
