The value of real estate is a fundamental part of the economy as it is used both as a collateral for loans as well as a key basis for a company’s annual reports. Today there are more than three million real estates in Sweden, and the market value of these can not be directly observed but have to rather be based on estimates. Thus, it is critical to attain reliable assessment and hire appraisers before making an investment decision or finalising an annual report. However, depending on the purpose and context of a valuation, the client may posses incentives to influence the outcome of the valuation in some direction. Such impact on the valuation of an asset is referred to as “client pressure,” and may come in various forms; expert pressure, information pressure and reward pressure. Does this have a real impact on market values? According to previous scholarly research, both domestic and international, client pressure is an observable phenomenon but its outcome is difficult to accurately assess. The purpose of this thesis is to provide evidence that supports the existence of client pressure within the commercial real estate market in Stockholm, and to establish a framework that may better understand the outcomes that such skewed valuations may lead to. In order to conduct this study, the authors have carried out a number of personal interviews with authorised appraisals, as well as further research into what can be done to prevent client pressure. The results of this study supports the notion that client pressure, primarily that of expert pressure, can be observed in the commercial real estate market in Stockholm. Moreover, what has been accentuated by the interviewed appraisals though is that such pressures do not affect the valuation outcome and that the discussion may actually provide a more accurate valuation as long as the appraisal manages to stay objective. However, several of the appraisals pointed out that they knew of others that have allowed themselves to be influenced. In order to counteract client pressure an authorization has been established by the organization Samhällsbyggarna - a development which has received great support by a large group of appraisals. Ultimately, the research in this study concludes that the best way to counteract client pressure is through knowledge, and our findings lead to the belief that experienced appraisals are willing to help less experienced ones if they are exposed to client pressure. / Värdet på fastigheter är en samhällsangelägenhet och gäller som bland annat säkerhet för lån och underlag för bolags årsresultat. I Sverige finns drygt tre miljoner fastigheter och marknadsvärdet på dessa är ingenting som kan observeras, bara bedömas. Därför har man behov av tillförlitliga bedömningar och anlitar värderingsmän inför såväl investeringsbeslut som årsredovisningar. Vad det bedömda värdet på fastigheten ska användas till kan ge incitament för kunden till att vilja påverka det åt något håll. Beroende på olika egenskaper hos kunden kan försök till att påverka värdet yttra sig i olika former utav press; expertpress, informationspress samt belöningspress. Kan det påverka bedömningen av marknadsvärdet? Pressen finns enligt tidigare forskning, det har undersökts både nationellt och internationellt, men konsekvenserna är svåra att avgöra. Studien i den här uppsatsen har fokuserats på att undersöka om press finns på Stockholmsmarknaden för kommersiella fastigheter och i vilken form den i så fall förekommer. För att ta reda på resultaten har ett antal intervjuer gjorts med aktiva värderare och det har också undersökts vad man i så fall gör för att motverka pressen. Det som kunnat konstateras är att kundpress finns och främst i form av expertpress. Värderarna menar dock att det inte påverkar deras egna bedömningar och att det till och med i vissa fall kan vara nyttigt att få en diskussion med kunden om de använda värdena så länge man fortsatt är objektiv. Flertalet har dock meddelat att de vet om fall där en värderare låtit sig påverkas. För att motverka kundpress finns auktorisering från organisationen Samhällsbyggarna och en stor del av värderarna tycker att det är ett bra stöd. Man motverkar även press med kunskap, erfarna värderare kan gå in och hjälpa de mer oerfarna med argument om man blir utsatt för press.

Real estate appraisal is the means of defining the market value of land and property affixed to it. Many different features determine the market value of a property. For example, the distance to the nearest park or the travel time to the central business district may be significant when determining its market value. The use of machine learning in real estate appraisal requires algorithm accuracy and interpretability. Related research often defines these two properties as a trade-off and suggests that more complex algorithms may outperform intrinsically interpretable algorithms. This study tests these claims by examining the impact of geospatial features on interpretable algorithms in real estate appraisal. The experiments use property transactions from Oslo, Norway, and adds relative and global geospatial features for all properties using geocoding and spherical distance calculations. Such as the distance to the nearest park or the city center. The experiment implements three intrinsically interpretable algorithms; a linear regression algorithm, a decision tree algorithm, and a RuleFit algorithm. For comparison, it also implements two artificial neural network algorithms as a baseline. This study measures the impact of geospatial features using the algorithm performance by the coefficient of determination and the mean absolute error for the algorithm without and with geospatial features. Then, the individual impact of each geospatial feature is measured using four feature importance measures; mean decrease impurity, input variable importance, mean decrease accuracy, and Shapley values. The statistically significant results show that geospatial features improve algorithm performance. The improvement of algorithm performance is not unique to interpretable algorithms but occurs for all algorithms. Furthermore, it shows that interpretable algorithms are not axiomatically inferior to the tested artificial neural network algorithms. The distance to the city center and a nearby hospital are, on average, the most important geospatial features. While important for algorithm performance, precisely what the geospatial features capture remains for future examination. / Fastighetsvärdering är ett sätt att bestämma marknadsvärdet på mark och egendom som anbringas på den. Flera olika variabler påverkar marknadsvärdet för en fastighet. Avståndet till närmaste park eller restiden till det centrala affärsdistriktet kan till exempel vara betydande när man bestämmer ett marknadsvärde. Användningen av maskininlärning vid fastighetsvärdering kräver noggrannhet och tolkbarhet hos algoritmer. Relaterad forskning definierar ofta dessa två egenskaper som en kompromiss och föreslår att mer komplexa algoritmer kan överträffa tolkbara algoritmer. Den här studien testar dessa påståenden genom att undersöka påverkan av geospatiala variabler på tolkbara algoritmer i fastighetsvärdering. Experimentet använder fastighetstransaktioner från Oslo i Norge, och lägger till relativa och globala geospatiala variabler för alla fastigheter med hjälp av geokodning och sfäriska avståndsberäkningar. Såsom avståndet till närmaste park eller stadens centrum. Experimentet implementerar tre tolkbara algoritmer; en linjär regressionsalgoritm, en beslutsträdalgoritm och en RuleFit-algoritm. Som jämförelse implementerar den också två artificiella neuronnätsalgoritmer som en baslinje. Studien mäter påverkan av geospatiala variabler med algoritmprestanda genom determinationskoefficienten och det genomsnittliga absoluta felet för algoritmen med och utan geospatiala variabler. Därefter mäts den individuella påverkan av varje geospatial variabel med hjälp av fyra mått på variabelbetydelse; mean decrease impurity, input variabel importance, mean decrease accuracy och Shapley-värden. De statistiskt signifikanta resultaten visar att geospatiala variabler förbättrar algoritmers prestanda. Förbättringen av algoritmprestanda är inte unik för tolkningsbara algoritmer utan sker för alla algoritmer. Dessutom visar resultatet att tolkningsbara algoritmer inte är sämre än de testade artificiella neuronnätsalgoritmerna. Avståndet till stadens centrum och det närmaste sjukhuset är i genomsnitt de viktigaste geospatiala variablerna. Även om de geospatial variablerna är viktiga för algoritmprestanda, kvarstår frågan om vad exakt de betyder för framtida granskning.

<p>As a mid-size industrial city in North America, the City of Hamilton has been increasingly experiencing urban sprawl in the past six decades coupled with population growth and economic development. The study of various interdependent processes driving the evolution of urban form requires the application of simulation models that offer urban planners and policy-makers an efficient means for evaluating urban development policies. This thesis focuses on the modeling efforts towards building a microsimulation residential housing market system for the City of Hamilton. To this end, two major tasks have been conducted in this research. First, a state-of-the-art agent-based microsimulation housing market framework has been designed. Second, two model components in the microsimulation framework, namely a real estate appraisal model and a new housing development model, have been estimated. The objective of the real estate appraisal model is to assess the market values of existing dwellings based on the housing transactions in the previous period. Thre e model forms, including a traditional hedonic model, a spatial regression model, and a regression Kriging model, have been employed in estimations for comparison purposes. A series of independent variables that describe the characteristics of dwelling, location, and neighborhood are specified in the explanatory model. The comparisons among estimation results demonstrate that the spatial regression model has achieved a higher goodness-of-fit than the traditional hedonic model. In addition, we verified that spatial autocorrelation is present in the residuals of the traditional hedonic model, which is explicitly captured by the spatial regression model. In terms of model prediction accuracy, spatial models (SAR and Kriging) both achieve a certain level of improvements over the traditional hedonic model. Overall, we end up recommending that the SAR model is more appropriate to be incorporated into the microsimulation framework, as it provides the best match between predicted and observed values. The new housing development model enables the development of a dynamic housing supply module in the simulation framework by modeling the location and type decisions during the housing development process for each year. A parcel -level two-tier nested-logit model has been estimated. The model is able to deal with not only the decision to develop a specific vacant residential land parcel, but also the development type choice. In terms of the factors influencing the decision to develop, the picture revealed from the model estimation results is that land developers are more likely to start a development project in greenfields than in brownfields. As for the type choice decision during the development process, a variety of variables describing transportation accessibility, residential amenities, the characteristics of the land parcel and neighborhood are included in the model specifications.</p> / Master of Arts (MA)

The integration of Vision Transformers (ViTs) using Masked Autoencoder pre-training (ViTMAE) into real estate valuation is investigated in this Master’s thesis, addressing the challenge of effectively analyzing visual information from real estate images. This integration aims to enhance the accuracy and efficiency of valuation, a task traditionally dependent on realtor expertise. The research involved developing a model that combines ViTMAE-extracted visual features from real estate images with traditional property data. Focusing on residential properties in Sweden, the study utilized a dataset of images and metadata from online real estate listings. An adapted ViTMAE model, accessed via the Hugging Face library, was trained on the dataset for feature extraction, which was then integrated with metadata to create a comprehensive multimodal valuation model. Results indicate that including ViTMAE-extracted image features improves prediction accuracy in real estate valuation models. The multimodal approach, merging visual and traditional metadata, improved accuracy over metadata-only models. This thesis contributes to real estate valuation by showcasing the potential of advanced image processing techniques in enhancing valuation models. It lays the groundwork for future research in more refined holistic valuation models, incorporating a wider range of factors beyond visual data. / Detta examensarbete undersöker integrationen av Vision Transformers (ViTs) med Masked Autoencoder pre-training (ViTMAE) i bostadsvärdering, genom att addressera utmaningen att effektivt analysera visuell information från bostadsannonser. Denna integration syftar till att förbättra noggrannheten och effektiviteten i fastighetsvärdering, en uppgift som traditionellt är beroende av en fysisk besiktning av mäklare. Arbetet innefattade utvecklingen av en modell som kombinerar bildinformation extraherad med ViTMAE från fastighetsbilder med traditionella fastighetsdata. Med fokus på bostadsfastigheter i Sverige använde studien en databas med bilder och metadata från bostadsannonser. Den anpassade ViTMAE-modellen, tillgänglig via Hugging Face-biblioteket, tränades på denna databas för extraktion av bildinformation, som sedan integrerades med metadata för att skapa en omfattande värderingsmodell. Resultaten indikerar att inklusion av ViTMAE-extraherad bildinformation förbättrar noggranheten av bostadssvärderingsmodeller. Den multimodala metoden, som kombinerar visuell och traditionell metadata, visade en förbättring i noggrannhet jämfört med modeller som endast använder metadata. Denna uppsats bidrar till bostadsvärdering genom att visa på potentialen hos avancerade bildanalys för att förbättra värderingsmodeller. Den lägger grunden för framtida forskning i mer raffinerade holistiska värderingsmodeller som inkluderar ett bredare spektrum av faktorer utöver visuell data.

近年來，由於經濟不景氣及市場上之不確定因素，使得投資不動產成為保險業資金投入的新興熱門標的。因投資不動產之行為涉及憲法所保障之財產權，如非有特殊事由不得加以限制。保險法第146條亦有規定不動產為保險業得投資之法定項目之一。惟於現行相關法規規範下，保險業投資發展空間受到限制。究監理機關於此所扮演之角色為何？監理目標為何？於現行法制下，相關法規是否妥適合理等問題，皆有待探求。又隨著不動產投資投入之資金日趨增加，保險業者受裁罰之案例亦較過去為多。究過去裁罰當中，主管機關之處分有無不妥適之處，以及保險業者於投資不動產之相關缺失中，是否有可改善之空間，為本文所欲探討之目標。 / In recent years, real estate has become an increasingly popular investment target for insurers’due to uncertainties in the market and the economic downturn. Legally, investing in real estate is a constitutional right to property and protected specifically by the Insurance Act, Article 146. It cannot be infringed upon unless for specific, legitimate reasons. However, current regulations put a fair amount of limitation on investment in real estate. The adequacy of these regulations, as well as the goals and roles of our supervision system, is to be discussed. In addition, there are a growing number of administrative sanctions as insurance companies put more and more funds in real estate. This article also attempts to investigate if these penalties were appropriate and if there’s room for improvement on the insurers’ part.

長期以來，不動產估價相關研究皆著重在估價理論與技術的改進，以使最後估值能更為精確。然而實際上，估價人員的行為才是影響價格評定的本質因素；尤其國內、外對於市場比較法的相關研究，皆發現估價人員有偏離理論規範的現象。故本文第一部份是分析估價人員系統行為偏誤問題；第二部分探討特徵價格模型的偏誤，並應用分量迴歸改進不同高低價格的估價模型準確度；第三部分比較估價師個別估價與大量估價模型之偏誤與差異；為提升客觀估價方法，第四部分探討以估價師估值樣本產生最可能價格與機率的方式。本研究藉由估價師估價行為與大量估價模型不同面向的研究，重新檢視市場比較法的各種偏誤現象。 一、市場比較法估價之系統行為偏誤 市場比較法是估價人員最常使用的方法，但在估價過程中主觀判斷部分相對較多也最易引起爭議。過去市場比較法研究著重在估價理論與技術改進，以減少主觀判斷造成的隨機偏誤，然而，研究指出行為偏誤造成的系統偏誤，其持續性的偏誤對於估價偏誤有重要影響。過去探討估價行為的研究，受限於資料樣本取得，幾乎都是以問卷或實驗得到結果，而非實證結果。本文則是以實際估價報告書進行實證分析，以鄒檢定(Chow Test)及似無相關檢定(Seemingly Unrelated Regression)，發現勘估標的估值與比較標的市場成交值兩模型係數有顯著差異，表示估價人員估值模型起始點高於市場成交值模型，估價人員對於使用類型有高估現象，對於面積估計有低估現象，支持估價人員有系統行為偏誤。此外，也發現評估不同市場時，行為偏誤程度亦不同，估價人員於評估拍賣價格時，因估價經驗與比較案例較少，較傾向參考比較案例，行為偏誤亦較小，僅面積係數有差異；但於評估正常市場價格時，估價行為偏誤較大，區位、使用類型及面積係數皆有差異。 二、不同高低價格不動產估價模型之偏誤改進－分量迴歸模型之應用 隨著國內不動產市場M型化推案趨勢，非典型住宅（如：高總價豪宅和低總價小套房）類型逐漸增多，對於此類型產品的估價精準度也需要提升。從過去研究發現，最小平方迴歸估計忽略各特徵屬性對價格條件分配的差異。本研究乃以分量迴歸方法建立住宅大量估價模型，藉以瞭解住宅特徵對於不同價格分量的差異，實證結果發現以最小平方迴歸模型估計相較於分量迴歸，對於一樓、頂樓、車位、區位等變數有高估或低估的情形。比較估值模型預測精確度，本文透過30次重複實驗，發現分量迴歸對於兩側尾端樣本有較佳的預測能力。從實證方法而言，本文改進以最小平方迴歸模型對兩尾端價格高估或低估問題；就實務應用方面，隨著不動產產品差異度增加，以及新版巴塞爾協定（Basel Ⅱ）實施對不動產價值更新的需求，分量迴歸模型可提升兩尾端估計精確度，並提供住宅大量估價系統另一種資產重估方法。 三、大量估價與個別估價之差異分析 不動產估價研究可分為個別估價與大量估價兩類，過去有關個別估價研究多為估價行為或估價方法改進，而大量估價研究則多是運用數量方法建立模型，由於兩者建構方式不同，欲同時進行兩者實證比較並不容易。本文藉由相同勘估標的，取得估價師個別估價資料，與資料庫建立的自動估價系統，進行案例選取、權重調整及最後估值三階段的比較分析。經本文差異比較發現，前二個階段的比較結果，皆是自動估價系統優於個別估價，從估價行為觀點，顯示自動估價系統具有較客觀且符合估價理論程序優點。然而第三階段的估值比較，發現自動估價系統的表現並不如預期，此結果與過去研究結論不同，分析兩者差異值大的案例，發現資料庫樣本較少地區以及非典型住宅類型（面積過大或過小），是造成個別估價估值與自動估價系統估值差異大的原因，顯示自動估價系統有其適用限制，未來若增加資料庫樣本或次市場模型，兩者估值差異將可獲得改善。 四、不動產估值認知與估值機率 本章以估價師估值資料為樣本，藉由蒙地卡羅模擬估值機率模型，其模擬結果可客觀估計最可能價格（the most probable price）與機率。由於資料樣本來自於專業估價師的判斷，估值分配代表的是估價師所估計的可能估值結果；相較於過去僅以單一估值決定資產價值，藉由多位專業估價師所判斷結果而建立的估值區間與最可能價格及機率，應該是更具客觀性且接近市場價值，而且估值機率的呈現將有助於估價報告書使用者評估存在於資產價值的上方（upside risk）及下方風險（downside risk）。 / In the real estate appraisal research, appraisers and mass appraisal are the two methods most often used in the sales comparison approach. Past studies focused on appraisers behavior and mass appraisal model, lack of compared the difference by appraisers for the same objects as well as mass. The first essay reviews the behavior literature regarding real estate appraisers and summarizes the two hypotheses for departure from normative models. The study is based on appraisers’ reports analyzes for the appraisal valuation and the market transaction by hedonic price models. Among the 112 appraisal reports of the transaction data of 230 comparables and the valuations of 224 subjects, the results reveal the effect of the variables on the appraisal valuation are not consistent with those of the market transaction. In addition, comparing the appraisers’ behavior on the general market with the auction market, the result found is the differences between the valuation models and the transaction models are less than the models on the auction market. The empirical evidences support the two hypotheses and can be explained plausibly by the appraisers’ behavioral contention. The second essay analize the current domestic trend of residential types, it is easy to find that high-priced dwelling units and low-priced dwelling units keep gaining popularity. Thus, the estimation of these two types of residences should be more precise. Since ordinary least square regression can not signify the variation caused by different quantile functions of a conditional distribution, this study estimates the housing price by quantile regression. We compare the models by using ordinary least square regression and quantile regression. The empirical results show that the distributions of some variables, such as first floor, top floor, parking lot, location, are different between two models. These are easily to be underestimated or overestimated when ordinary least square regression is applied. Based on thirty repeated experiments using random sampling, the results of hit rate and mean absolute percentage error show that quantile regression estimates more accurately on two-tailed distribution. For mass appraisal application, a quantile regression advances the estimate on two-tailed price and provides a new method on assets reevaluation of banks.

市場比較法估價過程需要經比較、分析及調整三階段，而估價師於個別估價應用中因缺乏標準化依據，造成估價過程常被質疑過於主觀且偏重經驗法則，導致估價結果產生因人而異現象。基於此本研究乃建立大量估價模型，運用估價師進行市場比較法行為邏輯，在選取比較案例階段採用明科斯基距離概念，並結合特徵價格理論，建立大量樣本的客觀標準化估價模型。藉由此模型分析，估價師不需要主觀預測，可改善過去估價結果不一致情形，並達到大量估價目的。 透過隨機抽樣的重覆實驗，實證結果發現，以模型的平均百分比預測誤差與命中率比較，整體而言未劃分次市場估價模型皆較劃分次市場準確，且其穩定度亦較高，而劃分次市場較不準確的原因，在於舊市區的表現不穩定，與市郊區的估值表現較差所致。經本研究結果，不論劃分次市場與否，標準化大量估價模型大致上均達到研究預設水準，然劃分次市場模型雖有時較準確但未優於未分區模型甚多，顯示適當劃分次市場可提高準確度，但資料太少時，劃分次市場可能產生統計偏誤問題。 / The process of the market comparative approach includes three stages：comparing ,analyzing and adjusting. Real estate appraisers lack the standardized basis in the individual appraisal application, so they are often challenged by subjectivity and stressing experience, which leads to the phenomenon that appraisal results are always different from person to person. Based on this, our research establishes the automated valuation methods. By applying the appraisers’ behavior in market comparative approach, we use minkowski metric in selecting comparative subject, and associate with hedonic price theory to establish objective standardized real estate appraisal model. By using this model, the appraisers can avoid subjective forecasting, reduce inconsistency, and therefore achieve the goal of mass appraisal. Through the repetition experiment of random sampling, we compare MAPE and Hit-rate between models. The result shows that non-delimited markets are more accurate and stable than delimited markets. The reason for this is the instability caused by older district of the city and bad performance in the suburban area. After these findings, no matter delimited market or not, the standardized mass appraisal model achieved the research standard we had set in advance. Sometimes the model of delimited market is more accurate, though not by a significant amount, it shows that higher accuracy can be expected through adequate market segmentation, but will lead to bias when lack of datum.

不動產估價一般需採比較法、收益法及成本法等三種方法查估。不動產估價最終估值決定須進行協調(reconciliation)，協調的目的係為完成關聯(correlation)之步驟，就各種方法資料之質量及優缺點進行分析。為使不動產估價對於比較法三件買賣實例，及三種方法估值採賦予權重之決定方式提供量化解釋，本文分別建立比較法內部及三種方法外部權重模型。內部權重部分，買賣實例（市場）比較法一般需蒐集數個比較標的，經調整後之試算價格決定比較價格。國外以數學計算式計算實例權重雖已相當普遍，但目前尚無應用特徵效用模式，解讀實例權重形成與比較標的內部條件之關聯。本文以2007年及2008年地價基準地6,435筆買賣實例建構特徵權重模型，發現比較法買賣實例權重受價格型態、交易日期接近性、是否屬近鄰地區、實例差異百分率絕對值加總、實例比較項目修正數、其他兩個實例相對值等自變數影響顯著。 欲探討成本法估值是否與成交價存在差異，以作為外部權重設定之基礎，本文以北部地區986筆交易案例，由估價人員逐筆以成本法估計成本價格，俾與成交價格比較。發現成本價格有高估之系統性偏誤現象，分量迴歸(quantile regression)分析實證認為成本法並未因屋齡較新之建物有較高精度。另發現房地交易價格愈高、建物單價愈高、總樓層數愈高、移轉樓層愈低、建物面積愈小、建築工期愈長及利潤率愈小者；估值差異愈小。 外部權重分為三部分，首先將估價過程中之諸項因子，以分析階層程序法(AHP)專家問卷，彙整各種方法權重因子；其次，基於最適加權平均模式在於使三種方法估值總誤差最小之觀點，經由數學計算方程式建構2,150筆基準地三種方法標準差及相關係數模型以計算權重。第三、為了瞭解比較估值、收益估值與土地開發分析估值之關聯，本文將2,150筆三種估價方式權重建立聯立模型，以三階段最小平方法(3SLS)進行估計。實證模型系統加權解釋力甚高，且三種方式權重之自變數多符合預期並顯著，顯見三種方式之關聯性。 最後，不動產估價仍需考慮一般因素，如金融海嘯對全球金融及房地產市場，其影響力無遠弗屆，最終估值之決定即需考慮市場景氣對最終估值之影響。為探討對於(不)景氣時是否(低)高估？影響(低)高估與否之影響因素為何？本文以2002年至2004年國內某金融機構對房屋貸款20,532件之估值，以二項式邏輯特(Binary Logit)模型分析。實證結果發現於不景氣時期抵押貸款低估機率提高，景氣時期則無高估現象。綜上，本文以權重模式、估值差異及市場景氣影響探討不動產估價最終估值之形成，於權重模型建構及預測上，非如以往文獻僅對估值預測，而係就權重預測。於加權平均應用上，外生變數之迴歸係數可作為權重設定之參考。本文直接探討最終估值形成之權重核心，冀使估價之客觀性及科學化程度提高。 / Real estate appraisal comprises the sales comparison, income, and cost approaches to value in general. The purpose of reconciliation is to complete the procedure of correlation and analyze the qualitative and quantitative strengths and weaknesses of different approach data. In order to assist quantifiable explanation when weighted average for three comparables in the Sales comparison approach and indicated values from three approaches are applied, this paper constructs internal and external weight models respectively. For internal weight model, this paper examines the correlation between weight and internal attributes of 6,345 sales comparables from land value benchmark in 2007 and 2008 adopting the hedonic price model. The outcome shows the price type, the proximity of transaction date, inside the neighborhood area or not, total gross adjustment as %, numbers of adjustments and the attributes of other two comparables considered in one appraisal are significant on the weight of comparables. To explore whether the cost approach causes bias or not, and make it reference for establishment of external weight model, this paper compares the cost value, appraised by valuers applying the cost approach individually, from a sample of 986 transactions of properties sold in 2007 and 2008 in northern Taiwan, to sale price and finds the cost value is higher than sale price on average. It proves that the reliability of the cost approach is comparatively questionable due to its systematic bias of overestimation. With quantile regression, the outcome shows that the precision of cost value won’t increase for newer buildings. In addition, this paper finds the more the total property sales amount, the higher the unit construction fee, the higher building, the lower story, the smaller area, the longer construction years of properties, and the smaller profit rate; the smaller the bias. There are three parts for external weights. First, AHP expert questionare is adopted to combine weight factor from each approach. Secondly, based on the logic that the best way to assign weights on three appraisal approaches is to get the minimum total error, this paper calculates the standard error and correlation indicators from three approaches using 2,150 land value benchmarks. Thirdly, in order to realize the weights correlation among the sales comparison value, income capitalization value and land development analysis value, this paper builds a model based on the three-stage least squares method simultaneous equation (3SLS). The empirical result shows system weighted R2 is high and most attributes on the weights of three indication of value are significant and are consistent with expected sign, which means the model fit is good and how the weights of three methods correlate. Finally, general factor also needs to be considered in real estate appraisal. For instance, financial tsunami exerts powerful influence on financial and real estate market globally, which makes it necessary to consider real estate cylce influence when seeking the final value. In order to discuss whether the appraisal value of mortgage is smaller (greater) or not when the market is bearish (bullish) and the corresponding factors, this paper collects 20,532 mortgage appraisal value from one bank from 2002 to 2004. With Binary Logit model, this paper finds the probability of lower appraisal is greater in bear market. The outcome confirms two hypotheses of this paper. However, overestimation is not confirmed in bull market. To sum up, this thesis researches the formation of final value of real estate appraisal by discussing weight model, appraisal bias and influence of real estate cycle. For weight model construction and forecasting, this dissertation forecasts weight directly, instead of value like most literature focus. The regression coefficients estimated from factors during the procedure of each approach could serve for reference if weighted average is applied reconciling the value conclusion by valuers. By delving into the core issue of value formation, it hopes to elevate the degree the objectivity and science of real estate appraisal.

The Thesis deals with impact of land price on the usual price of building and its change due to code no. 89/2012. The impact was determined based on market prices of lands and buildings retrieved from real estate databases. From statistically processed data was found out a 75% correlation between the market price of building and land. For numerical description the impact of land price on the usual price of building was assembled an equation. With this equation can be from land price roughly estimated usual price of building. The results of this thesis can contribute to speeding-up the process of initial estimation of usual price, as well as presents a way of displaying location, price, distribution of data and frequency of equal prices in one graph.

估價於房地產市場實扮演著一不可或缺的角色，精確的估價不僅可提供消費者正確極充分的購屋資訊，亦為政府擬定政策方針之基礎。由於台灣房地產市場為一不完全市場，消費者在購屋的同時更常因資訊的不健全而遭受不必要之損失，因此精確及流通之估價資訊實為健全台灣房地產市場之首務。 鑑於過去估價技術仍未成熟，所佔之房價常無法令人信服。本研究欲以類神經網路之功能，將其原理應用於房地產估價上，試圖解決過去估價方法本身之缺失，並作為估價人員輔助之工具。本研究主要以倒傳遞及理解倒傳遞類神經網路與特徵價格法進行公證比較分析，並以特徵價理論為基礎，利用類神經網路得出影率房地產價格更具代表性之因素，以做為未來建立房地產估價輔助系統之參考。 為了解不同的資料型態是否會使類神經網路有不同的學習效果，本研究將資料分為四組實驗設計，分別對不同的資料型態進行測試，研究結果顯示類神經網路對於資料型態較為敏感，其中又以理解倒傳遞類神經網路為最，使得其在預測能力上易受異常點或極端值的影響，而有好壞差異較大的情況。即類神經網路之學習效果端視資料是否具代表性而定。