基于凝聚的层次聚类算法的改进  

Improved Algorithm Based on Agglomerative of Hierarchical Clustering

作　　者：石剑飞[1] 闫怀志[2] 牛占云[1] 

机构地区：[1]北京理工大学计算机网络攻防对抗技术实验室,北京100081 [2]北京理工大学软件学院,北京100081

出　　处：《北京理工大学学报》2008年第1期66-69,共4页Transactions of Beijing Institute of Technology

基　　金：国家部委基础科研项目(20021823)

摘　　要：为提高基于凝聚的层次聚类算法的准确率.在研究了空间级约束适用情况的基础上,以Single Link算法为例,验证了空间级约束条件对聚类结果的影响.与实例级约束Single Link算法相比,空间级约束Single Link算法只需较少约束条件即可达到较高准确率.实验结果证明,空间级约束可以有效提高聚类的准确率.To enhance the accuracy of hierarchical clustering algorithm, based on a study of spaceconstrained application, and using the Single Link algorithm as an example, the impact of spaceconstrainment is certificated to the result. Compared with the case-constrain Single Link algorithm, space-constrained Single Link algorithm needs less constrained conditions to achieve higher accuracy. Experiments showed the space-constrainment can enhance the clustering accuracy effectively.

关 键 词：聚类算法 实例级约束 空间级约柬 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]