搜索
3962JOURNALOFHOUSINGSTUDIES,VOLUME18NO.1,JUNE2009(本文於2008年9月30日收稿，2008年11月23日審查通過，實際出版日期2009年6月)***64Ph.D.,NationalChengchiUniversity,No.64,Sec.2,ZhiNanRd.,Taipei.E-mail:yinghui@nccu.edu.tw*AComparisonofAutomatedRealEstateValuationModelsandAppraisers’Appraisals—ComparablesSelection,WeightingAdjustmentandValuationfortheSalesComparisonApproach***Ying-HuiChiang**摘 要()關鍵詞：不動產估價、大量估價、自動估價系統、比較法AbstrActInrealestateappraisalresearch,appraisersandmassappraisalsarethetwomethodsmostoftenusedinthesalescomparisonapproach.Inthepast,mostappraisers’studiesfocusedontheappraisers’behaviororonimprovingappraisingmethods.Sincethesetwomethodsareconstructedindifferentways,itisnoteasytofindenoughobjectsfortheanalysisandacomparisonbetweenappraisersandmassappraisal.Themainpurposeofthispaperistoselectcertaintargetcasesandthencomparethedataprovidedbyappraisersaswellastheresultsofmassappraisals.Theanalysisisconductedinthreestages:selectingcomparables,adjustingweightsandestimatingvalues.ThedifferencesintheanalysisindicatethatthedifferencesintheAutomatedValuationModel(AVM)oftheMinkowskimetricofthecomparablesissmallerthantheappraisalsprovidedbyindividualappraisers.ThefindingsindicatethattheapplicationoftheAVMmodelismoreobjectiveandfitstheappraisalprocedure.However,withthelimitsoftheAVM,thesparsenessofthedataandnon-typicalhousing,theresultsoftheAVMinsomeareasdifferfromthosemadebyappraisers,whichindicatesthattheAVMhaslimitations.Thissituationisexpectedtoimproveasaresultofaccumulatingmoredataandadjustingthemodels.IftheaccuracyoftheAVMisassured,theAVMwillbeappliedmorewidely.Keywords:realestateappraisal,massappraisal,automatedvaluationmodel,salescomparisonapproach40(automatedvaluationmodel,AVM)1970(Carbone&Longini,1977)(FannieMae)(FreddieMac)(Valentine,1999)(Ross&Nattagh,1996,Moore,2005)(multipleregressionmodel)(spatialregressionmodel)(neutralnetworkmodel)(artificialintelligencemodel)Ibrahimetal.(2005)(1)Fabozzi(1998)Valentine(1999)&Waller(1999)(TheAppraisalInstitute)(2)(estimate)(opinion)(anappraisal)(2007)Fisher(2002)(2007)Dell(2004)()(3)(2007)41(1.2.3.)(Minkowskimetric)Detweiler(1999)Ibrahimetal.(2005)Moore(2005)(2007)(2007)(semi-parametric)Pace(1995)Bin(2004)Clapp(2004)(2007)(artificialneuralnetwork)McGrealetal.(1998)Wongetal.(2002)(1998)(2007)Detweiler(1999)Moore(2005)(computerassistedmassassessment,CAMA)AEP(feedbackmodel)(transportablecost-specifiedmarket,TCM)AEP3042(pairwisedifferences)(coefficientofdeviation)(InternationalAssociationofAssessingOffice,IAAO)(meanabsolutepercentageerror,MAPE)(4)(hit-rate)(5)(MAPE)(hit-rate)(2005)(2006)()(2007)43Fisher(2002)(BaselII)(6)()(2007)(2006)(一)資料來源(7)2004120052(8)3,13210%90%AVM44(2007)53.6%(9)9.4%30(10)2004120052(二)自動估價系統建立之設計AVM(2007)βPagourtzietal.(2003)(11)Todora&Whiterell(2002)(12)(三)自動估價系統之準確性(hit-rate)(MAPE)AVMCalhoun(2001)PricewaterhouseCooper4%73%48%(MAPE)8.1%20.9%9.9%Jonas(1990)4510%30%20%67%Matysiak&Wang(1995)10%30%20%70%(2007)MAPE15.64%25.49%10%33.73%43.82%20%62.28%74.56%(2007)MAPE20.41%24.48%(FE)5%-15%28%26%15%58%60%(2007)MAPE17.97%10%43%20%63%15.48%10%20%41.82%71.65%(四)個別估價樣本資料表一 國內相關文獻大量估價模型準確性比較表Hit-rateMAPE10%20%(2007)()43.82%74.56%14.45%()33.73%62.28%19.56%()()40.28%71.33%15.64%()36.09%64.39%25.49%()(2007)()43%63%17.97%(2007)()41.82%71.65%15.48%305%-15%15%(2007)266224.65%295420.20%266024.48%285820.41%460(13)30200412005652.1532231110(一)自動估價系統與個別估價比較案例選取之差異比較(1)Minkowskimetric∑wi[abs(χsiχci)/χsi]...................................................................................(1)wiiχsi(χci)()i(01)()()(14)(01)1(Wi)(β)(Wi)(14)47表二 個別估價勘估標的樣本資料說明表123395.684224170223395.684224170340623.3122134314402716.9922114325402718.16221343165544.15322418074543.03322415089645.8932219099645.89322190109645.89322290119646.793224901223368.844224103213563130.254224220141439.053224140151428.653221140161022.512211231171422.512212231181022.512211231191022.512212231201016.281111140211039.28111414022420.12221123123420.12221123124439.05322414025428.6532211402616849.332241402716866322414028563142.85422122029563132.22422119030563132.224221219052.152.801.931.632.8011.0310.4038.690.890.250.492.226.5114.9648表三 自動估價系統與個別估價選取之比較標的明科斯基距離比較表10.160.1420.250.140.190.020.230.02120.3517.4191.1417.4130.010.0240.090.070.000.010.020.0114.1023.1218.2719.0450.100.0361.990.160.030.011.060.0325.6626.2653.1017.0270.420.2180.040.220.230.030.050.0955.0812.91123.5339.5690.050.22102.180.170.030.090.460.0655.0039.5621.2536.48110.920.13120.050.010.370.030.020.0040.3123.3147.1222.02130.810.20140.140.010.380.020.040.0047.4712.3829.2331.15150.110.01160.010.010.010.000.010.018.0231.1576.0776.07170.030.05180.020.080.000.010.010.003.7617.0251.271.43190.020.06200.160.300.010.010.040.0856.8616.5621.9026.79210.020.13220.090.080.020.020.040.00119.8618.0948.205.57230.020.08240.020.130.010.000.030.0278.655.57141.2417.78250.070.01264.740.150.010.000.030.039.664.110.6717.11272.640.17283.950.292.270.030.010.0986.1115.120.1730.98293.080.30303.620.172.610.093.420.0184.9030.7594.504.4312(CV)CV100%S49()301963.33%4(451416)16795%(0.12491.2970)73表四 成對母體平均數