臺灣博碩士論文加值系統

研究者常利用顯示性偏好來建立住宅不連續選擇多項Logit模型（MNL），然而顯示性偏好資料通常只表示觀察樣本實際發生或經驗過的選擇行為屬性值，卻缺少未體驗方案的屬性值。為了使多項Logit模型的校估更精確，透過設定的方式來組成未體驗方案的屬性值，舉例來說，能透過可觀察變數的平均值取代未知的資料，但無論怎樣組成資料，都將造成校估與建立多項Logit模型時的誤差，並可將其視為一種系統的不確定性。為了避免這樣的不確定性，本研究提出倒傳遞類神經網路模型。倒傳遞類神經網路模型可以估計家戶實際的選擇，透過受訪者實際選擇行為的經驗，以啟發式的過程來達到學習的目標。本研究將焦點在方法論的比較上，並試圖證明多項Logit模型在系統不確定條件下的誤差。此外，結合顯示性偏好資料與新的分析工具來達到預估住宅選擇行為的可能。

本研究資料來源取自中央研究院經濟研究所於民國88年起所建立「華人家庭動態資料庫（Panel Study of Family Dynamics, PSFD）」。基於此資料，本研究使用三種設定方式來建立多項Logit模型，並探討不同的屬性值對多項Logit模型的影響。此外，本研究透過資料百分比變化的處理來探討模型的穩定性。而在倒傳遞類神經網路模型方面，則是透過學習演算法的變化來檢驗模型的效率與精確度。

在進行比較多項Logit模型與倒傳遞類神經網路模型與假說驗證後，發現不依賴假設資料的倒傳遞類神經網路模型有較高的精確率，其實證結果摘要如下：(1)對於家戶未體驗方案的屬性值，其設定方式會造成統計意義上各式各樣的MNL模型。(2)MNL在模型的配適度上較不受資料比例的影響，而是對屬性設定的方式較為敏感。(3)MNL模型對於家戶遺失資料的設定方式的差異影響模型的配適度，亦影響了預測準確率。(4)MNL模型在變數的顯著性與參數符號上，依舊受到資料比例的影響，單從 的變化較不易觀察。(5)依照本研究所訓練的網路結構的條件下，BFG演算法所建立的模型是屬於效率與穩定度兼具的模型。(6)AAE、MAPE與FE<5%的指標中，BPN所建立的模型平均來說都較MNL模型穩定且誤差率也較低。

故利用顯示偏好資料做為住宅選擇行為模型時，傳統分析工具MNL對於方案未體驗效用的屬性，透過不同的設定方式會改變模型的變數符號與顯著性，導致模型校估上的偏誤，進而影響模型的穩定性與準確率。整體而言，在各項的評估指標中顯示，兩種分析工具皆會受到建模樣本比例不同程度的影響，仍以不使用假設性資料的BPN模型在模型穩定度與預測精度上較佳。
關鍵字：顯示性偏好、模糊不確定、多項Logit模型、多層前饋式倒傳遞神經網路模型

It has been usually using revealed preference data to establish housing discrete choice of multinomial logit models (MNL). However, revealed preference data normally only show alternatives that observed samples actually happen or experience and leave unobserved attribution variables missing. To help MNL calibration practically, those missing data are made up in one way or another, for example, using the mean of observed variables instead. That will generate bias on calibrating and estimating MNL model and can be considered one kind of system uncertain. To avoid the uncertain, Back-Propagation Neural Network (BPN) models is suggested. The BPN model can estimate household actual choice by using heuristic process to approach learning goal through household choice behavior of the actual experience. This paper is focus on comparative methodology, and trying to prove the bias of MNL under the system uncertain conditions. Besides, combining the revealed preference data with the new method will make estimate household choice be possible.
The study used the data of Panel Study of Family Dynamics (PSFD) executed in 1999 by the Institute of Economics for empirical tests. Base on the data, this paper use three kinds of setting method to establish MNL models and explore the effects of different character value for MNL model. Besides, this paper processes the variance on the data percentage in order to discuss the effect on the model stability. Nevertheless, BPN model is processed through a variety of different learning algorithms to examine the model of efficiency and accuracy.
Comparing MNL models with BPN models and examining the hypothesis, the BPN models has higher accuracy that not dependent on assume data. The result are summarized as follows: 1. The different setting of household missing data does have various MNL models that are statistically significant.2. Different sampling methods would not affect the value, but is more sensitive to the different data setting way.3. The different setting of missing data is not only change the value, but also affect MNL model stability and accuracy.4. Different data percentage would lead to the bias on the model of the dynamics of the parameter sign and the calibrated model.5. The BFG model has stability and accuracy under this paper to establish BPN model conditions.6. The BPN model has more stability and accuracy than MNL model in the AAE、MAPE and FE index.

Keyword：Revealed Preference (RP), Uncertain assumption, Multinomial Logit (MNL),Back-Propagation Neural Network (BPN)

第一章、緒論............................................................................................... 1
第一節 研究動機 ......................................................................................... 1
第二節 研究目的 ......................................................................................... 3
第三節 研究對象與範圍 ............................................................................. 4
第四節 研究內容與流程 ............................................................................. 4
第二章、Logit 模型 ..................................................................................... 9
第一節 個體選擇理論 ................................................................................. 9
第二節 Logit 模型的偏誤 .......................................................................... 14
第三節 多項羅吉特模型的理論架構 ....................................................... 17
第四節 羅吉特模型的統計特性 ............................................................... 20
第五節 名詞定義 ....................................................................................... 22
第六節 小結 ............................................................................................... 22
第三章、類神經網路................................................................................. 24
第一節 類神經網路理論 ........................................................................... 24
第二節 倒傳遞類神經網路理論 ............................................................... 29
第三節 類神經網路應用 ........................................................................... 34
第四節 類神經網路應用於住宅相關研究 ............................................... 39
第五節 小結 ............................................................................................... 40
第四章、實驗設計與資料分析................................................................. 42
第一節 實證設計 ....................................................................................... 42
第二節 實證資料說明 ............................................................................... 43
第三節 變數定義與說明 ........................................................................... 44
第四節 資料分配型式與方案集合說明 ................................................... 46
第五節 資料結構說明 ............................................................................... 47
第六節 評估指標 ....................................................................................... 48
第七節 敘述統計分析 ............................................................................... 49
第五章、實證結果..................................................................................... 56
第一節 資料分配80% MNL 模型的比較 .................................................. 56
第二節 資料分配50% MNL 模型的比較 .................................................. 62
第三節 資料分配30% MNL 模型的比較 .................................................. 68
第四節 MNL 模型小結 .............................................................................. 74
第五節 比較不同分配的BPN 模型 ........................................................... 75
第六節 綜合評析 ....................................................................................... 80
第六章、結論與建議................................................................................. 89
第一節 結論 ............................................................................................... 89
第二節 建議 ............................................................................................... 91

一、國內文獻
左心強2002《應用模糊理論於房屋選購查詢系統》碩士論文，私立元智大學。
江濤、陳小鳴 2006〈基於BP神經網路的住宅需求預測〉《武漢理工大學學報-信息與管理工程版》28(9):001-003.
吳至中2005 《檢視住宅選擇行為中的異質性與相似性─以台灣地區都市層級選擇為例》碩士論文，國立成功大學。
吳志文2002《以類神經網路探討船舶事故之研究》碩士論文，國立成功大學。
吳義斌 2006 《都市住宅發展程度與空屋推估之關係研究─以台中市為例》碩士論文，私立逢甲大學。
呂孟學 2002 《應用類神經網路於即時停車需求預測之研究》碩士論文，國立中央大學。
林祖嘉 2002 〈台灣地區餘屋存量影響因素之分析〉《住宅學報》 11:23-34。
林祖嘉、林素菁 1994 〈台灣地區住宅需求價格彈性與所得彈性之估計〉《住宅學報》4: 25-48.。
林素菁、林祖嘉 2001 〈台灣地區住宅供給彈性之估計〉《住宅學報》 10: 17-27。
袁淑湄 2003 《應用混合Logit模型探討台灣家戶住宅選擇之研究》碩士論文，國立成功大學。
張春龍 2000 《住宅選擇模型之研究─以台南市為例》碩士論文，私立長榮大學。
張雅惠 2002 《租賃住宅之需求彈性與所得彈性》碩士論文，國立成功大學。
連經宇 2003 《應用模糊語意方法與不連續選擇理論建立家戶購屋選擇行為模式之研究》碩士論文，國立成功大學。
陳建良、林祖嘉 1998 〈財富效果、所得效果與住宅需求〉《住宅學報》 7: 83-99。
陳彥仲 1996 《從個體不連續選擇行為探討臺灣地區住宅需求之價格彈性及所得彈性》行政院國家科學委員會專案研究計畫。
陳彥仲、薛立敏 1999《台灣地區出租住宅市場之研究》台北：內政部建研所。
陳啟賢 2005《影響高雄市住宅供需因素之研究》碩士論文，國立中山大學。
陳淑美、張金鶚 1998〈戶長及其配偶對住宅區為選擇之影響力分析〉《住宅學報》7：59-81。
陳獻文 2003 《以巢式Logit方法解析程序性住宅選擇模式─以新竹市為例》碩士論文，中華大學.
傅美生1984 《台北都會區住宅權屬與住宅類型選擇之研究—Logit模型應用》碩士論文，中興大學。
曾喜鵬 2004 《居住遷移與住宅選擇模式建立之研究--以住宅消費失衡與調整為觀點的解釋》碩士論文，國立台北大學。
華昌宜1996〈台北市住宅政策總體檢問題與建議〉《住宅學報》 4: 89-103。
黃舜卿 2008 〈我國消費者信心指數與民間消費支出之研究〉 《Economic Research》8：P121-158
賈宜鳳1984 〈台北都會區住宅需求函數之估測〉《台灣銀行季刊》35(3)：270-290。
劉惠雯1985《台北都會區社區/住宅選擇之研》碩士論文，中興大學。
蔡宜哲 1996 《從個體不連續選擇行為探討台灣地區住宅需求之價格彈性及所得彈性分析》碩士論文，成功大學。
蔡國委2002《台南市區域房地產預售屋量與人口遷移關係之探討》碩士論文，私立長榮大學。
蔡瑞煌、高明志、張金鶚1999〈類神經網路應用於房地產估價之研究〉《住宅學報》8:001-020。
鄧建民1985《臺灣地區住宅需求之研究》碩士論文，中興大學。
賴碧瑩、陳治勳 2007〈模糊類神經網路於不動產價格預測之研究〉《中華民國都市計劃學會論文研討會》 P44.
魏玉好 1993《台北房屋市場之研究》碩士論文，東吳大學。
二、國外文獻
Agrawal, D. and C. Schorling 1996 "Market share forecasting: An empirical comparison of artificial neural networks and multinomial logit model", Journal of Retailing 72(4): 383-407.
Aleksander I, Morton H 1990 "An Introduction to Neural Computing", Chapman & Hall, London.
Barney, D. K. and J. A. Alse 2001 "Predicting LDC debt rescheduling:Performance evaluation of OLS, logit, and neural network models", Journal of Forecasting 20(8): 603-615.
Bhat, C. R 1998 "Accommodating flexible substitution patterns in multi-dimensional choice modeling: Formulation and application to travel mode and departure time choice." Transportation Research Part B-Methodological 32(7): 455-466.
Bhat, C.R. 2001 " Quasi-random maximum simulated likelihood estimation of themixed multinomial logit model", Transportation Research 35B(7): 677-695
Börsch-Supan A1987 " Econometric Analysis of Discrete Choice: With Applicationon the Demand for Housing in the U.S. and West-Germany", Berlin,Heidelberg: Springer-Verlag
Bryan C. Pijanowsk, B. A. S., Marvin E. B, Kali E. S 2001 "Using GIS Artificial Neural Networks And Remote Sensing To Model Urban Change In The Minneapolis-st.Paul And Detroit Metropolitan Areas." ASPRS Proceedings 2001, St. Louis, Mo.
Cantarella, G. E. and S. de Luca 2005 " Multilayer feedforward networks for transportation mode choice analysis: An analysis and a comparison with random utility models" 121-155.
Celikoglu, H. B 2006 " Application of radial basis function and generalized regression neural networks in non-linear utility function specification for travel mode choice modelling", Mathematical and Computer Modelling 44(7-8): 640-658.
Cheng, C. B., C. L. Chen, et al 2006 " Financial distress prediction by a radial basis function network with logit analysis learning", Computers & Mathematics with Applications 51(3-4): 579-588.
Cripps, N. N. a. A 2001 " Predicting Housing Value-A Comparison of Multiple Regression Analysis and Artificial Neural Networks", JRER 22(3).
D. Martin 1997 " Early warnings of bank failures", Journal of Finance 52 (3)：853-868.
de Palma, A., N. Picard, et al 2006 " Discrete choice models with capacity constraints: An empirical analysis of the housing market of the greater Paris region", Conference in Honor of the Retirement of Kenneth A Small, Irvine, CA.
Feng Shan, L. D. X 1997 " An integrated knowledge-based system for urban planning decision support", Knowledge-based system 10: 103-109.
Fischer, M. M. and S. Gopal 1994 " ARTIFICIAL NEURAL NETWORKS - A NEW APPROACH TO MODELING INTERREGIONAL TELECOMMUNICATION FLOWS", Journal of Regional Science 34(4): 503-527.
Follain, J. R 1982 " Does inflation affect real behavior – the case of hosing." Southern
Economic Journal 48(3): 570-582.
Garcia, J. A. B. and J. E. R. Hernandez 2007 " Housing and urban location decisions in Spain: An econometric analysis with unobserved heterogeneity", Urban Studies 44(9): 1657-1676.
Gillingham, R. and Hangemann, R 1983 " Cross-Sectional Estimation of a Simultaneous Model of Tenure Choice and Housing Services Demand", Journal of Urban Econmics,Vol.14,PP.16-39.
Hensher, D. A. and T. T. Ton 2000 " A comparison of the predictive potential of artificial neural networks and nested legit models for commuter mode choice".
J.H. Aldrich and F.D.Nelson 1984 " Linear Probability, Logit and Probit Models", Sage Publications,Beverly Hills,CA.
Kain, J. F. and Quigly, J. M 1975 " Housing Market and Racial Discrimination: A Microeconomic Analysis”,the National Bureau of Economic Research, New York.
Kim, J. H., F. Pagliara, et al 2005 " The intention to move and residential location choice behavior " UrbanStudies 42(9): 1621-1636.
Kumar, Akhil, Vithala R. Rao and Harsh Soni 1995 " An Empirical Comparison of Neural Network and Logistic Regression Models", Marketing Letters, 6(4) 251-263.
Li, M. M 1977 " A Logit Model of Homeownership", Economerica,Vol.45,No.5, PP.1081-1091.
Makridakis, Spyros, Steven C. Wheelwright and Victor E. McGee 1983 " Forecasting: Methods andApplications", New York: John Wiley and Sons.
Margarita M. Lenk, E. M. W. a. A. S 1996 " High-tech valuation: should artificial neural networks bypass the human value? " Property Valuation & Investment 15(1): 8-26.
Mayo, S. K 2002 " Theory and Estimation in the Economics of Housing Demand"
Mohan, S. V. I. a. B. K 2002 " Urban Landuse Monitoring Using Neural Network Classification." IEEE.
NAG, B. A. J. A. B. N 1997 " Performance Evaluation of Neural Network Decision Models ", Management information System. 14(2): 201-216.
Nijkamp, P., A. Reggiani, et al 1997 " Spatial choice behaviour: logit models and neural network analysis." Annals of Regional Science 31(4): 411-429.
O. M. Lewis, J. A. W. a. D. H. J 2001 " Identification of Residential Property Sub-Markets using Evolutionary and Neural Computing Techniques", Neural Comput & Applic 10 108-119.
Park, S. R. H. a. S. Y 1996 " RBF-Neural Network Based Land Cover Classification Technique of Satellite Image for Pollutant Load Estimation."
Paul J. Roebber, M. R. B. a. S. J. R., Thomas J. Grafenauer 2006 " Realtime forecasting of snowfall using a neural network "
Quigley, J. M. 1985 " Consumer Choice of Dwelling, Neighborhood and Public Service", Regional Science and Urban Economics, Vol.15,PP.41-63.
R. O’Ryan, L. L., F.J. Martínez " Neural Networks for the Analysis of Urban Decontamination Policies: an Application to Santiago."
Rossini, P. K. a. P. 1999 " Using Neural Networks to Estimate Constant Quality House Price Indices", Fifth Annual Pacific-Rim Real Estate Society Conference Kuala Lumpur, Malaysia: 26-30.
Schintler L, Olurotimi OO .1997 " The Use of Neural Networks in Planning for Demand-Responsive Transit, in: Himanen V, Nijkamp P, Reggiani A (eds.) ", Neural Networks in Transport Systems. Avebury, Aldershot (forthcoming).
Seko, M. and K. Sumita 2007 " Japanese housing tenure choice and welfare implications after the revision of the tenant protection law" Journal of Real EstateFinance and Economics 35(3): 357-383.
Sobol, I. M. 1998 "On quasi-Monte Carlo integrations".
Srinivasan, K. K. and H. S. Mahmassani 2003 " Analyzing heterogeneity and unobserved structural effects in route-switching behavior under ATIS: a dynamic　kernel logit formulation." Transportation Research Part B-Methodological 37(9): 　793-814.
Stanley McGreal, A. A., Dylan McBurney and David Patterson 1997 " Neural networks: the prediction of residential values", Journal of Property Valuation & Investment, 16(1): 57-70.
Thurston, J. 2002 " GIS & Artificial Neural Networks Does Your GIS Think."
van Campen,C. and E. van Gameren 2005 " Eligibility for long-term care in The Netherlands: development of a decision support system." Health & Social Care in　the Community 13(4): 287-296.
Walker, J. and M. Ben-Akiva 2002　" Generalized random utility model"
Weeks, M. 2003 " Discrete choice methods with simulation", Journal of Applied 　
Econometrics 18(3): 379-383.
X. Dai, S. K. 1999 " Data fusion using artificial neural networks: a case study on 　multitemporal change analysis", Computers, Environment and Urban Systems 23: 19-31.
Xia, A. G.-o. Y. L. 2004 " Integration of Neural Networks and Cellular Automata for Urban Planning", Geo-spatial Information Science (Quarterly) 7(1).
Yates, J. and D. F. Mackay 2006 " Discrete choice modelling of urban housing markets: A critical review and an application." Urban Studies 43(3): 559-581.

三、參考書籍
李茂能2006《結構方程模式軟體AMOS之簡介及其在測驗編制上之應用》心理出 版社股份有限公司。
凌瑞賢 2001《運輸規劃原理與實務》[初版] 鼎漢國際工程顧問股份限公司。
葉怡成 1999《類神經網路應用與實作》[第六版]，儒林圖書有限公司。