在傳統銀行放款的觀念中認為，借款人主導了還款的來源，關於貸款風險的研究大多集中在借款人行為因素的探討，但是房屋貸款的風險，除了借款人本身的特質外，應該還需要不同角度的探討，尤其是在擔保品方面。銀行在辦理放款時，對擔保品價值的評估僅以當時的市場價值作直接的判斷，並依判斷結果來決定貸款的成數，這樣的決策並未考慮擔保品本身所處的區域條件及其未來的發展性，因而產生了風險判斷的偏誤。 本研究將透過不動產的供需價量的關係，嘗試找出影響房屋貸款擔保品風險的因子，並對房屋貸款的風險因子給予適當的權重及評分，再運用劃分等級的模型，將研究區域依房屋貸款風險的大小劃分風險等級。最後利用不同的角度或方法檢驗各種模型對區域風險分類之異同及功能，以建立模型提供銀行於承做房屋貸款或制定放款政策時，作為決定貸款成數(LTV)的參考依據，避免銀行貸款日後遭受擔保品價格下跌所產生的風險。 本研究以分析層級程序法(AHP)及分析網路程序法(ANP)設計不同的問卷，在取得各風險因子的權重後，對各項風險因子時間序列的數據進行分析，最終取得台北縣市各區域的風險等級。實證結果AHP及ANP皆通過一致性分析，AHP與未權重化ANP間不具顯著差異；權重化ANP與極限化ANP間不具顯著差異。AHP權重與ANP未權重化矩陣兩種模型在區域房屋貸款風險等級的區分標準上較為寬鬆。ANP權重化矩陣及ANP極限化矩陣對區域房屋貸款風險等級的劃分較為嚴格。這兩類不同等級劃分標準的模型提供金融業者在制定房屋貸款政策時可以有多樣的選擇。 / People with traditional concept of bank lending believe that borrowers dominate the sources of repayment. Researches regarding to the credit risk of loan mainly focus on the behaviors of borrowers. Nevertheless, the risk of mortgage loan should be deliberated with different points of view, especially the collateral, besides considering the characteristics of borrowers. During the process of loan, banks evaluate the collateral with directly determine according to the prevailing market value and decide the proportion of loan. The decision is not considered with the regional factors and the future development of the collateral. A bias of risk determination therefore exists during the process. The research tries to find the factors that influence the collateral risk of mortgage loan through supply, demand, price and quantity of real estate. Also, it allocates the weight and evaluation of every risk factor of mortgage loan. The research then distinguishes the investigated areas into different risk levels according to the mortgage loan risk by applying appropriate model. The research stands at various points of view and utilizes different methods to determine whether the classification of area risk is appropriate or not and offering a banks model to be the reference basis of determining the loan to value(LTV) when executing mortgage loan or drawing up loan policy. Banks can avoid the risk of collateral depreciation in the future. The research designed various questionnaires with Analytic Hierarchy Process (AHP) and Analytic Network Process (ANP). After obtained the weights of different risk factors, an analysis was processed on time sequence data of every risk factor and results the risk level of every subarea of Taipei. The empirical results in consistency analysis by AHP and ANP are passed. The difference between AHP and un-weighted ANP is not significant. The difference between weighted ANP and limited ANP is also not significant. Both weighted AHP and un-weighted ANP matrix models are lenient on the classify criteria of area mortgage risk levels. In opposition, weighted ANP matrix models and ANP limited matrix are strict on the same criteria. The two models with different criteria offer financial corporations different choices when drawing up policies of mortgage loan