金融機構為增加放款市佔率，紛紛提高額度、降低利率及鬆綁條件，然而卻也同時提高呆帳的風險。個案公司為台灣租賃業龍頭（車貸為其重要業務）為研究對象。貸款績效的良窳在於對逾放的管理，因此發展風險評估模式，降低風險對個案公司甚至對整個租賃業，都是重要的議題。本研究邀請專家確認影響風險評估之主要因素，也由於汽車貸款風險評估可視為灰色系統問題，因此進一步以灰關聯分析為基礎發展評估模式。對評估模式而言，特異度代表實際呆帳資料中，被判斷為呆帳的比例，因此特異度愈高的預測模式對個案公司而言更具實用性。所發展模型與其他資料探勘技術相較下，的確具有較佳的整體正確率與特異度，因此顯示出所發展模型於風險評估之優越性，亦完全符合租賃業的實務需求。此一模式可有效輔助個案公司於車貸之審查，並可簡化現有審查流程，降低呆帳風險，達成永續經營的目標。

Financial institutions have raised loan quota, declined interest rates, and loosened conditions to increase the market share. However, this leads to a high risk of bad debt expense. The company we study is a leader for the Taiwan leasing industry, and automobile loan is its important business. The key of loan performance lies in the management of overdue repayment, so developing risk assessment model and reducing loan risk are important issues for the case company and even the whole leasing industry. This study invited experts to identify the main factors affecting risk assessment, and because the risk assessment of automobile loans is a grey system problem, the grey relational analysis forms the basis for developing the evaluation model. The specificity is the proportion of people who test bad debt among all those who actually do have bad debt. This means that the prediction model with high specificity is more useful for the company we study. Compared with other data mining techniques, since the results show that the proposed model has higher accuracy and specificity, it not only shows the superiority of the proposed model for risk assessment, but completely satisfies the practical requirements for the leasing industry. The proposed model has the benefit of simplifying the current review process by effectively reviewing car loan. By reducing the risk of bad debts, it can help the company we study to achieve the goal of sustainable development.