摘要
从自检的角度对进化K-means聚类算法进行了改进,在分裂后通过评价函数评价聚类结果,保证正确的分裂能够连续进行,同时对不正确的分裂进行阻止.在UCI数据库中的Iris和Wine数据集上进行实验,验证了引入自检策略的进化K-means算法比进化K-means算法优越.
Evolutionary K-means(F-EA) is sensitive to variation rate on the determination of cluster forMutation, it can cause cluster for Mutation and split direction changed frequently, weaken the splittingeffect and reduce the clustering quality. An improved evolutionary K-means algorithm was designedfrom the standpoint of self-inspection, this algorithm evaluated clustering results after splitting byevaluation function, ensuring the correct division can be continued while blocking the incorrectdivision. Two experiments were planned in the Iris and Wine data sets from the standard UCIrepository, the experimental results prove that the improved evolutionary K-means algorithm issuperior to the evolutionary K-means algorithm.
出处
《东北师大学报(自然科学版)》
CAS
CSCD
北大核心
2014年第3期59-63,共5页
Journal of Northeast Normal University(Natural Science Edition)
基金
吉林省自然科学基金资助项目(20130101060JC)
吉林省教育厅"十二五"科学技术研究项目(2014132
2014125)
关键词
进化K-means
分裂
评价函数
自检
evolutionary K-means
split
evaluation function
self-inspection