不同以往使用單期模型估計違約機率，本研究採用離散時間Hazard模型，並利用多期Logistic模型估計參數，最後與單期模型進行樣本內外預測準確性之比較。
資料方面，經完整性處理過後，篩選出共85,932家企業資料 (共1,082,493筆季資料) 並利用離散時間Hazard模型，納入以下五大構面變數共22項變數:（1）總體經濟因素（2）負責人基本資料（3）企業特性（4）財務比率以及（5）企業授信資料，用以評估違約機率，並利用模型特性，達到即時動態調整違約風險，同時找出影響違約機率之顯著變數。
結論為下，在1%信心水準下，顯著影響違約與否的有授信額度、實收資本額、員工人數、負責人性別、負責人有無配偶、擔任本企業人年限。總體變數中，國內生產毛額成長率、五大銀行放款利率變動率、躉售物價指數變動率、失業率、信義房價指數變動率，都顯著影響違約機率。產業變數方面，食品業為違約機率最高的產業。樣本外預測，Hazard 模型正確率達94.86%優於單期模型正確率93.75%。樣本內預測，Hazard模型正確率達82.08%，同樣也優於單期之80.85%。

Different from the past which the single-period model is used to estimate the probability of default. this study uses the discrete-time hazard model and utilizes the multi-period logistics model to estimate the parameters. Ultimately, the single-period model is compared with Hazard model in term of accuracy of out-sample data testing and in-sample data testing.
After data cleaning processing, the total of 85,932 corporate data (the total of 1,082,493 quarter-firm data) were screened. Five major facet including Macroeconomic factors, Basic information of the owner, Enterprise characteristics, Financial ratio and corporate credit information were incorporated and used to assess the probability of default. By the properties of Hazard model, we can achieve real-time dynamic adjustment of default risk and identify significant variables at the same time.
The conclusion is that the amount of paid-in capital, the number of employees, the gender of the owner, whether the owner has a spouse, and year of work in the same industry are strongly related to bankruptcy. Among the Macroeconomic variables, the rate of change of gross domestic product (GDP), the interest rate of the five major banks, wholesale price index (WPI), the unemployment rate, and Xinyi house price index all significantly affect the probability of default. In terms of industrial, the food industry is the industry with the highest probability of default. According this research found, it indicates that the Hazard model has a correct rate of 94.86% on out-of-sample testing, which is better than the single-stage model with a correct rate of 93.75% and the Hazard model has a correct rate of 82.08% on in-sample testing, which is better than the single-phase 80.85% as well.

摘要...........................i
Abstract......................ii
誌謝辭........................iii
目錄..........................iv
表目錄........................vi
圖目錄.........................vi
壹.緒論........................1
貳.文獻探討....................3
參.研究方法....................8
肆.實證分析...................13
伍.模型正確性與樣本外預測......19
陸.結論與建議.................24
柒.參考文獻...................26

(1). 陳家彬等(2003)。商業銀行對企業授信決策考量因素與授信品質之關係。管理評論，22(2)，1-23。
(2). 吳懿娟(2004)。我國貨幣政策傳遞機制之實證分析。中央銀行季刊， 26(4)，33-68。
(3). 蔡明春、鄭青展、馮淑鈴、陳坤志(2009)。中小企業貸款違約預警模式—以某商業銀行為例。文大商管學報，14(1)，19-40。
(4). 張瑞娟與欉清全（2009），貨幣政策與匯率轉嫁之探討─臺灣之實證分析，應用經濟論文，86，頁31-67。
(5). 陳昇鴻、鍾國貴、周隆耀(民102)。台灣地區中小企業內部授信評等的決定因素：企業主特質與財務狀況、企業財務特性、產業生命週期與景氣展望的角色。台灣金融財務季刊，14(2)，41-70。
(6). 詹博翔(2016)。銀行對中小企業授信評估因子與授信風險關係之探討。東吳大學企業管理研究所碩士論文，台北。
(7). Berkson, J. (1944). Application of the logistic function to bioassay. Journal of American Statistical Assodation, 39, 357 -65.
(8). Okun, Arthur M.“Potential GNP: Its Measurement and Significance.”In Proceedings of the Business and Economics Statistics Section. Alexandria, VA: American Statistical Association, 1962, pp. 98-103.
(9). Beaver, W. H. (1966), “Financial ratios as predictors of failure,” Journal of Accounting Research 4, 71-102
(10). Altman, E. I. (1968). Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy. The Journal of Finance, 23(4), 589-609.
(11). Altman, E., R.G. Haldeman, and P. Narayanan, (1977), “ A New Model to Identify Bankruptcy Risk of Corporations.” Journal of Banking and Finance, pp.29-54.
(12). Ohlson, J. A., 1980. Financial Ratios and the Probability Prediction of Bankruptcy. Journal of Accounting Research, 18(1): 109-131.
(13). Shumway, T. 2001, Forecasting Bankruptcy More Accurately: A Simple Hazard Model. Journal of Business 74:101-124.