一种影响空间下的快速K-means聚类算法.pdfVIP

小型微型计算机系统 ??年?碌?期 Vol??37 ????9?? ?啪? ? ???? ?唧?? Systems 一种影晌空间下的快速?猰??聚类算法 赵文冲，蔡江辉，赵旭俊，张继福 ???萍即笱Ъ扑慊?蒲в爰际跹г海?????? ?猰??簔?????????甤? 摘要：尽???是一种经典的迭代划分聚类算法，但在每次迭代过程中，需要计算和比较每个数据点与所有中心点之间的距 离．因此聚类过程时间开销大．利用影响空间数据结构，给出一种快速尽???聚类算法．该算法首先，引入影响空间数据结构 ??????????????????????????????????????????????????????????????????????????????me??s???????????????????????????? 类别即是该区域中所有数据点所属的类别，有效地降低迭代过程中的数据量，提高了聚类效率；最后，理论分析和实验结果表 明，仅对代表性数据点而非数据集中所有数据点进行迭代聚类，能够在保证聚类质量的前提下，有效地提高聚类效率． 关键词：聚类；影响空间；区域划分；代表元素 ????????????1??11????????????A 文章编号：???????????掰 F??t??mea??Cl??te??ngAlgoritl??B??ed? ???? Space ?? ??猚??，? JiaIlg??hui????A0xu-jun??ZHANG?咖 (School ? ???? Science ?? 1ktulology??Tai????uIlive????of????瞕 ??????琓?”柚???。??? ??糟?：忌??? ? a classic ?? ??撕? ???? ????? ????．??豶??? needs ? calculate and compafe the dist??ces be??een?? ?? ??? ?? all cluster ??蒭? ? ?? iteration??wllichm ?簁? a huge??ecost??Ttlispaper ???腶 ??尽??? algorit??????sny??algori??m?? ???? ????籥 ??? ?? ?鯿??—? part??tion??e?? set ? ?? small ?舀???鷇 ?孙 ?? a ?????? ?? ???忧? ?鷕???? ?? correlalion re??on??tllenmnniIlg tlle??me??salgodtllm ?廿? ??????№ set until??terionfIlnction conVe??ence??neclass ? a represent??object? also??eclass ? all ?? ??? iD??ec??relation??gion?? The pmposed ???廿? wi??lower?锄?? ? ????? process ? iIIlprove the clus??gef!fidency??Tk??cal?????謉 ?． ????? results ?? t}lat ????? not on??e? da??adbjects in??l??set??but?吐? repfesentilIg ?? set not oIlly ??総? the ????? accuracy??butoI