預(yù)測(cè)電信行業(yè)客戶流失——基于一種sas生存分析模式的應(yīng)用程序【外文翻譯】

1、標(biāo)題：PredictingCustomerChurnintheTelecommunicationsIndustry––AnApplicationofSurvivalAnalysisModelingUsingSAS原文：ABSTRACTConventionalstatisticalmethods(e.g.logisticsregressiondecisiontreeetc.)areverysuccessfulinpredictingcus

2、tomerchurn.Howeverthesemethodscouldhardlypredictwhencustomerswillchurnhowlongthecustomerswillstaywith.Thegoalofthisstudyistoapplysurvivalanalysistechniquestopredictcustomerchurnbyusingdatafromatelecommunicationscompany.T

3、hisstudywillhelptelecommunicationscompaniesunderstcustomerchurnriskcustomerchurnhazardinatimingmannerbypredictingwhichcustomerwillchurnwhentheywillchurn.Thefindingsfromthisstudyarehelpfulftelecommunicationscompaniestoopt

4、imizetheircustomerretentiontreatmentresourcesintheirchurnreductioneffts.INTRODUCTIONInthetelecommunicationindustrycustomersareabletochooseamongmultipleserviceprovidersactivelyexercisetheirrightsofswitchingfromoneservicep

5、rovidertoanother.Inthisfiercelycompetitivemarketcustomersdemtailedproductsbetterservicesatlesspriceswhileserviceprovidersconstantlyfocusonacquisitionsastheirbusinessgoals.Giventhefactthatthetelecommunicationsindustryexpe

6、riencesanaverageof3035percentannualchurnrateitcosts510timesmetorecruitanewcustomerthantoretainanexistingonecustomerretentionhasnowbecomeevenmeimptantthancustomeracquisition.Fmanyincumbentoperatsretaininghighprofitablecus

7、tomersisthenumberonebusinesspain.Manytelecommunicationscompaniesdeployretentionstrategiesinsynchronizingprogramsprocessestokeepcustomerslongerbyprovidingthemwithtailedproductsservices.Withretentionstrategiesinplacemanyco

8、mpaniesstarttoincludechurnreductionasoneoftheirbusinessgoals.Indertosuppttelecommunicationscompaniesmanagechurnreductionnotonlydoweneedtopredictwhichcustomersareathighriskofchurnbutalsowemonthlybillsareconsideredinthestu

9、dy.Highvaluecustomersarethesewithmonthlyaveragerevenueof$Xmefthelastthreemonths.Ifacustomer’sfirstinvoicecoverslessthan30daysofservicethenthecustomermonthlyrevenueispratedtoafullmonth’srevenue.Granularity–Thisstudyexamin

10、escustomerchurnattheaccountlevel.Exclusions–Thisstudydoesnotdistinguishinternationalcustomersfromdomesticcustomers.Howeveritisdesirabletoinvestigateinternationalcustomerchurnseparatelyfromdomesticcustomerchurninthefuture

11、.Alsothisstudydoesnotincludeemployeeaccountssincechurnfemployeeaccountsisnotofaproblemaninterestfthecompany.SURVIVALANALYSISCUSTOMERCHURNSurvivalanalysisisaclanofstatisticalmethodsfstudyingtheoccurrencetimingofevents.Fro

12、mthebeginningsurvivalanalysiswasdesignedflongitudinaldataontheoccurrenceofevents.Keepingtrackofcustomerchurnisagoodexampleofsurvivaldata.Survivaldatahavetwocommonfeaturesthataredifficulttohlewithconventionalstatisticalme

13、thods:censingtimedependentcovariates.Generallysurvivalfunctionhazardfunctionareusedtodescribethestatusofcustomersurvivalduringthetenureofobservation.Thesurvivalfunctiongivestheprobabilityofsurvivingbeyondacertaintimepoin

14、tt.Howeverthehazardfunctiondescribestheriskofevent(inthiscasecustomerchurn)inanintervaltimeaftertimetconditionalonthecustomeralreadysurvivedtotimet.Therefethehazardfunctionismeintuitivetouseinsurvivalanalysisbecauseitatt

15、emptstoquantifytheinstantaneousriskthatcustomerchurnwilltakeplaceattimetgiventhatthecustomeralreadysurvivedtotimet.Fsurvivalanalysisthebestobservationplanisprospective.Webeginobservingasetofcustomersatsomewelldefinedpoin

16、toftime(calledtheiginoftime)thenfollowthemfsomesubstantialperiodoftimerecdingthetimesatwhichcustomerchurnsoccur.It’snotnecessarythateverycustomerexperiencechurn(customerswhoareyettoexperiencechurnarecalledcensedcaseswhil