资源描述:
《airbnb分析报告》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库。
1、ZhiweiWangAirbnb:NewUserBookingsAbstractInsteadofwakingupinahotel,Airbnballowstravelerstobooklocalpeople’shousefortheirtrips.Givenuser’sdataofAirbnb(USusersonly),thetaskistopredictwhichcountryanewuser'sfirstbookingdestinationwillbe.Theinitialresponsevariablehas12destina
2、tioncountries.However,duetolimitedinformationandimbalanceofdataset,thetaskismodifiedtobedistinguishingwhetheranewuser’sfirstdestinationwillbeinUS,othercountriesordonottravelatall.Thisreportcomparesdecisiontree,randomforestandSVMwithdifferentkernels.Italsoshowshowtodealw
3、ithimbalanceddata,missingvalues/outliersandvariablesindateformat.Comparingtesterrorrate,SVMmethodsdobetterthantreebasedmethods.Consideringefficiency,SVMwithlinearkerneltakesmuchshortertimethatotherkernelsbutdecisiontreecertainlyrequiresleastcalculatingresourcesamongallm
4、ethods.IntroductionPredictinganewuser'sfirstbookingdestinationisquietimportantforAirbnbcompany.ImaginethatrightafteryousignuponAirbnb,withoutdoinganysearch,thewebsitewillpostyourfavoritecountryonitshomepage.WealreadyknowthatAmazonrecommendsitemstoyoubasedonyourbuyingand
5、searchingrecords.However,thistimeAirbnbwillonlyuse“irrelevant”informationsuchasuser’sage,browser,ordevicetypetodothisprediction.Amazonrecommendsitemsbasedonuser’sbuyingandsearchinghistory.Thisprojectraisesadifferentquestion:Isitpossibletopredictanewuser’sfirstchoiceofde
6、stinationonlybasedonhissignupinformation?SometimespeoplesignuponAirbnb,thenjustclosethewebsiteandsoonforgetit.NowifAirbnbwantstosendadvertisementemailtothoseusers,itwishestoknowwhichcountry’spromoinformationorcouponsshouldtheemailsinclude.Someinitialguess:Peoplewhoconne
7、ctstohisaccountfromphoneisnotlikelytobookaninternationaltrip.OldpeoplearemorelikelytotravelwithinUS.Peoplewhodonotcompletetheiruserinformationperhapswon’ttravelatall.DatasetdescriptionResponsevariable:Thereare12possibleoutcomesofthedestinationcountry:'US','FR','CA','GB'
8、,'ES','IT','PT','NL','DE','AU','NDF'(nodestinationfound),and'other'.'NDF'isdifferentfrom'other'because'other'm