如果您无法下载资料，请参考说明：
1、部分资料下载需要金币，请确保您的账户上有足够的金币
2、已购买过的文档，再次下载不重复扣费
3、资料包下载后请先用软件解压，在使用对应软件打开
OwenandAnil/MahoutinActionLastsaved:SAVEDATE\@"M/d/yyyy"\*MERGEFORMAT5/28/2010Lastsaved:SAVEDATE\@"M/d/yyyy"\*MERGEFORMAT5/28/2010OwenandAnil/MahoutinAction©ManningPublicationsCo.PleasepostcommentsorcorrectionstotheAuthorOnlineforum:http://www.manning-sandbox.com/forum.jspa?forumID=XYZ©ManningPublicationsCo.PleasepostcommentsorcorrectionstotheAuthorOnlineforum:http://www.manning-sandbox.com/forum.jspa?forumID=XYZPAGEXXX7IntroductiontoClusteringThischaptercovers:Ahands-onlookatClusteringinactionUnderstandingthenotionofsimilarityRunningasimpleclusteringexampleinMahoutThevariousdistancemeasuresusedforclusteringBirdsofafeatherflocktogether.Ashumanbeings,wetendtoassociatewithlike-mindedpeople.Wehaveagreatmentalabilityforfindingrepeatingpatterns,andwecontinuallyassociatewhatwesee,hear,smellortastetothingsthatarealreadyinourmemory.Forexample,thetasteofhoneyremindsusmoreofthetasteofsugarthansalt.Sowegrouptogetherthethingsthattastelikesugarandhoneyandcallthem“sweet”.Withoutevenknowingwhat“sweet”tasteslike,weknowthatallthesugarythingsintheworldaresimilarandofthesamegroup.However,weknowhowdifferenttheyarefromallthethingsbelongingtothesaltygroup.Unconsciously,wegrouptogethertastesintosuch“clusters”.So,innaturewehaveclustersofsugarythingsandsaltythings,witheachgrouphavinghundredsofitemsinit.Innature,weobservemanyothertypesofgroups.Considerapesversusmonkeys,whicharebothkindsofprimates.Allmonkeyssharesometraitslikeshortheight,longtail,andflatnose.Ontheotherhand,apesarecharacterizedbytheirlargesize,longarms,andbiggerhead.Apeslookdifferentfrommonkeys,butbotharefondofbananas.Soitisentirelyuptoustothinkofapesandmonkeysastwodifferentgroups,orasasinglegroupofbanana-lovingprimates.Therefore,whatweconsiderasaclusterentirelydependsonthetraitswechooseformeasuringthesimilaritybetweenitems(inthiscase,primates).Sowhatistheprocessofclusteringallabout?Supposeyouweregiventhekeystoalibrarycontainingthousandsofbooks.However,inthislibrarythebooksarearrangedinnoparticular
