文档介绍：大连理工大学学位论文独创性声明作者郑重声明:所呈交的学位论文,是本人在导师的指导下进行研究工作所取得的成果。尽我所知,除文中已经注明引用内容和致谢的地方外, 本论文不包含其他个人或集体已经发表的研究成果,也不包含其他已申请学位或其他用途使用过的成果。与我一同工作的同志对本研究所做的贡献均已在论文中做了明确的说明并表示了谢意。若有不实之处,本人愿意承担相关法律责任。学位论文题目:7<询j良雾乒看礅趔左用作者签名: 聋髫日期:i竖年上月乙日 k均值算法研究及其应用 Research andApplication ofK—means algorithm Ab stract 111edata mining technology was born when information explosion toobtain implicit relationships andpatterns from thelarge— them,clustering analysis isan important accident datasets、Ⅳith cluster analysis is a relatively new application thecontinuous accumulation oftraffic accident data,there are more requirements using datamining technique tofindnew accidentpatterns,which Caneffectively avoid Cluster algorithms especially K-means algorithm are a better choice. Firstly,we have anin—depth study inK--means clustering algorithm is awidely used clusteringalgorithm,which isknown ashighefficiency and , thealgorithm itselfexistdrawbacks inmany areas,including theimproper selection ofthe initialcluster center point and outliers seriously disturb clustering improved fortheK·-means algorithm which based on density--weighted may has a better clusteringresult;however,plexity ofthealgorithm is0(N2),which isveryinefficient. Thispaper proposes animproved,both toget betterclusteringefficient,and alsoensuresthat theimproved algorithm oflinear ,we provide amethod for the selection oftheinitialcluster center point,which Can selectthe center point inthe dataobject paper alsogives thecorresponding detecfion methods toprevent outliers disturbing theclustering theoreticalproofandexperiments,we prove thatthe improved algorithm ismore efficient thenK-means algorithm based on density-weighted. Inthispaper,we haveanin—depthstudy inthe characteristics ident data. 劢e ident data types often contain continuous values and discrete values,The existing clusteringana