財務危機預測 (Financial crisis prediction)係資料探勘技術之重要應用領域。當企業發生財務危機，對股東、債權人、員工及銀行等，都會造成極大的衝擊。而財務危機可能是肇因於企業本身的經營不善，例如： 擴張過速、投資過多、巨額負債、財務調度失敗等眾多因素。因此如何綜合考量這些因素，建立財務預警模型，提供正確的訊息，讓投資大眾提早察覺企業的危機，以調整決策，減少損失，是值得關心的課題。
為了強化授信及資產品質，金融機構利用企業及客戶的個人及財務資料，以決策樹、羅吉斯迴歸、類神經網路等演算法建構財務危機預測及信用評分模型，並透過財務危機預測及信用評分模型對企業及客戶進行評分以利區別企業及客戶之財務風險。
過去文獻致力於建構較高正確率之混合式方法，然而，有鑒於經濟因素之多元與複雜性及其影響因素之不確定性，造成財務危機之預測益趨複雜與困難，此外，礙於不易執行函數組合與難以解釋決策模型所隱含的資訊，致使銀行風險管理模型之實務應用亦受到侷限。
雖然有不同種類的統計方法可被用來預測未來財務危機發生的可能性，許多研究顯示類神經網路的預測能力勝過傳統的統計方法。但因現有經濟因素之複雜性及其影響因素之不確定性，造成財務危機的預測是非常困難的情況。由於類神經網路具有執行非線性統計模型之能力，並可從實例中直接學習不需要瞭解系統的數學模型為何，而直接以神經網路取代系統的模型，一樣可以得到輸入與輸出之間的關係。因此，自從人工智慧和類神經網路快速發展後，這些方法便取而代之的被廣泛運用在財務危機預測上。
此外，由於接收者操作特徵 (Receiver operating characteristic) 不需要明確定義非結構性假設，以及金融機構好客戶與壞客戶潛在分配之特性，本研究結合接收者操作特徵方法與類神經網路於企業財務危機預測。本建置方法採用接收者操作特徵方法及獨立樣本t檢定 (independent-sample t test)來選取適合之預測變數。此外，由於類神經網路具有較佳之財務危機預測能力。因此，本研究使用接收者操作特徵方法所選取之變數，並運用類神經網路來預測企業財務危機。希冀藉由建構之混合式方法來提昇企業財務危機預測之正確率，並降低投資損失。
本研究以1998-2014發生財務危機之臺灣觀光產業資料進行測試，用以評估混合式財務危機預測方法與其他三個方法之效率及性能差異。實驗結果發現混合式財務危機預測方法較其他三個建置模型，有較高之平均正確預測率；此外發現混合式財務危機預測方法較其他四個建置模型，有較低之型I及型II誤差。因此，本研究所建構之混合式財務危機預測方法能提供企業決策單位作為早期財務危機預警的參考，並提昇企業決策之品質及管理控制。
此外，為了讓金融機構在授信戶發生違約前，儘早採取防範或補救措施，以減少金融機構之呆帳損失。第二部份本研究利用新的變數選取方法、客群區隔 (Customer segmentation)以及評分方式，建構新的二維度評分方法 (Dual-scoring method)，期提升信用評分模型的效能，並應用於銀行之客戶挽留及催收策略。本研究蒐集銀行既有房貸戶之內部資料及該授信戶於金融聯合徵信中心之個人信用資料，並針對上述兩類型資料進行資料前處理及產生候選模型的變數選取。接著本研究利用基因演算法適應函數分析影響房貸戶繳款狀態特徵資訊之輸入變數，然後將各種變數之隱藏特徵之規則轉換為重要性的值。運用基因演算法排序變數之重要性能系統化辨識變數之有用性，其有助於模型輸入變數之選取，特別是排除不重要之輸入變數保留重要輸入變數。然後應用分類迴歸樹依據客戶特徵將客戶分類成同質性的風險客群，進而提高模型之預測力。本研究並分別針對不同客群建立單維度的信用評分模型，並校準多個評分模型以衍生出一致性評分與勝算比間之關連性。最後整合房貸行為評分模型及徵信局評分模型建構二維度評分方法，目的是要使用目標和特定的資料去預測客戶繳款之可能性。
我們也將本架構的二維度評分方法與單維度評分模型加以比較，並且以台灣某金融機構真實資料進行一系列的實驗，來評估二維度評分方法的效率及性能。實驗結果顯示二維度評分方法透過額外的風險評估和區隔，更進一步確認房貸資產組合內需要加強管理或控管的地方；此外發現二維度評分方法之預測力均顯著高於單維度行為評分模型及徵信局評分模型。

Financial crisis prediction is an important application field of data mining. The financial crisis will have a great impact on shareholders, creditors, employees and banks. It may be attributed to the mismanagement of the enterprise itself, such as expanding too fast, excessive investment, huge debt, financial failures, etc. It is a topic worthy of concern to develop early warning models for predicting financial crisis. The financial early-warning model can provide the correct information which can be aware of the financial crisis in the early stagy, such that the better decision-making can be made and the bad debt losses can be reduced.
Some financial institutions employ the personal information and the financial data of institutional customers and individual customers to enhance credit and asset quality management. They utilize the decision tree, the logistic regression (LR), and the neural network algorithm to construct financial crisis prediction. Then the financial institutions use financial crisis prediction to score the institutional customers and individual customers in order to distinguish the good credit institutional customers and individual customers from the bad ones.
Previous studies have focused on developing more accurate classifiers with various hybrid architectures. Due to the complexity of the existing economic factors and its influence factors of uncertainty, it is very difficult to forecast the financial crisis. Furthermore, research on the practical application of combined classifiers is limited owing to the difficulty of implementing functional composition or explaining the underlying principle behind decisions when applying the hybrid approach for risk management.

While there are different kinds of statistical approaches that can be exploited to forecast the likelihood of future financial crisis, many studies indicate that the NNs have better predicative ability comparing to the traditional statistical approaches. By reason of the complexity of the existing factors, the forecast of financial crisis is an intricate issue. The neural networks (NNs) present their capabilities to implement nonlinear analytical modeling that endow with the new alternative to other analytical techniques and to learn forthrightly from examples devoid of requiring or providing a methodical resolution to the problem. As a result, the NNs evolve expeditiously and become a significant alternative measures in recent years.
Besides, because the receiver operating characteristic (ROC) do not need to specify the structural assumptions and the underlying distributions of the outcomes for the good and bad credit institutional customers, this paper combines a ROC method and the NNs for business financial crisis prediction. The proposed approach adopts the ROC curve analysis and independent-sample t test for selecting adequate variables in prediction. Because of its better financial crisis prediction performance, the NNs are used to predict business financial crisis with the selected variables in ROC. The proposed hybrid approach employs the combined model for improving accuracies of business financial crisis prediction and limiting investment loss.
The firms in tourism industry of Taiwan that incurred financial crisis in 1998-2014 are selected for test. We make experiments using the collected data for comparing the prediction accuracy of the proposed hybrid approach with those of the other three famous methods. The experimental results reveal that the proposed approach can obtain better prediction accuracy than those of the others. Furthermore, we can find that the proposed approach outperforms the other three methods in the comparison of Type I and Type II errors. From these experimental results, we realized that the proposed hybrid approach provides an early-warning system of business financial crisis, which can promote decision making quality and management control.
In addition, in order to let financial institutions prior to the credit default, take precautionary or remedial measures as soon as possible, in order to reduce bad debt losses of financial institutions. The second part of this study proposed a dual-scoring method based on developing a new feature selection, customer segmentation, and scoring procedure in order to improve the accuracy and efficiency of credit scoring models. Furthermore, this study applied the dual-scoring method in the on-lending retaining and collection actions. This study collects bank’s internal data and credit bureau data from the Public Credit Registers (PCR) on existing mortgage customers, and engage in data preprocessing of feature selection from candidate models for the above two types of data. After data preprocessing, this study exploits the nature of the genetic algorithms (GAs) fitness function to analyze the input variables that influence mortgage payment status for feature information, and then converts the rules of the hidden features of the various variables and transforms them into importance values. Utilizing the GAs to rank the importance of variables enables systematic identification of their usefulness, which is helpful in model input selection, specifically for eliminating ineffective inputs while preserving useful ones. After feature selection, this study applies the CART to classify customers into homogeneous risk groups based on the customer characteristics, and thereby enhance the predictability of the model. This study builds a one-dimensional credit scoring model respectively for segmented groups and calibrates several scoring models to derive a consistent score-to-odds relationship after customer segmentation. Finally, this process combines the mortgage credit scoring model with the credit bureau scoring model to construct a dual-scoring method. The purpose is to predict the possibility of customer payment using objective and specific data.
We have also compared the dual-scoring method of this framework with the one-dimensional credit scoring models, and used real-life application datasets of a financial institution to conduct a series of experiments to evaluate the efficiency and performance of these algorithms. The experiment results show that the dual-scoring method undergoes additional risk assessment and segmentation to further identify the parts requiring enhanced management or control within a mortgage portfolio. Besides, the predictive ability of the dual-scoring method is clearly higher than the one-dimensional behavioral and credit bureau scoring models.

[1] J. Akhavein, W. S. Frame, and L. J. White, “The diffusion of financial innovations: An examination of the adoption of small business credit scoring by large banking organisations,” The Journal of Business, Vol. 78, No. 2, pp. 577-596, 2005.

[2] S. Akkoc, “An empirical comparison of conventional techniques, neural networks and the three stage hybrid Adaptive Neuro Fuzzy Inference System (ANFIS) model for credit scoring analysis: The case of Turkish credit card data,” European Journal of Operational Research, Vol. 222, No. 1, pp. 168-178, 2012.

[3] E. I. Altman, “Financial Ratio, Discriminant Analysis and Prediction of Corporate Bankruptcy,” Journal of Finance, Vol. 23, No. 4, pp. 589-609, 1968.

[4] E. I. Altman, G. Marco, and F. Varetto, “Corporate distress diagnosis: Comparisons using linear discriminant analysis and neural networks,” Journal of Banking and Finance, Vol. 18, No. 3, pp. 505-529, 1994.

[5] B. O. Antonio, I. D. Ana, O. A. Maria, and W. Nicholas, “Improving Bankruptcy Prediction in Micro-Entities by Using Nonlinear Effects and Non-Financial Variables,” Czech Journal of Economics and Finance, Vol. 65, No. 2, pp. 144-166, 2015.

[6] S. Asadi, R. Hashemi, B. J. Navid, and A. Arfai, “The validation of real customers and their credit risk management in banks using logit model,” Technical Journal of Engineering and Applied Sciences, Vol. 3, No. 24, pp. 3576-3583, 2013.

[7] A. F. Atiya, “Bankruptcy prediction for credit risk using neural networks: a survey and new results,” IEEE Transactions on Neural Networks, Vol. 12, No. 4, pp. 929-935, 2001.

[8] A. D. Aydin, and S. C. Cavdar, “Prediction of Financial Crisis with Artificial Neural Network: An Empirical Analysis on Turkey,” International Journal of Financial Research, Vol. 6, No. 4, pp. 36‐45, 2015.

[9] A. Blanco, R. Pino-Mejías, J. Lara, and S. Rayo, “Credit scoring models for the microfinance industry using neural networks: Evidence from Peru,” Expert Systems with Applications, Vol. 40, No. 1, pp. 356-364, 2013.

[10] K. K. Boyer, and G. T. M. Hult, “Customer behavior in an online ordering application: a decision scoring model,” Decision Sciences, Vol. 36, No. 4, pp. 569-598, 2005.

[11] F. Castillo, K. Marshall, J. Green, and A. Kordon, “A methodology for combining symbolic regression and design of experiments to improve empirical model building,” Genetic and Evolutionary Computation Conference, pp. 1975-1985, 2003.

[12] F. L. Chen, and F. C. Li, “Combination of feature selection approaches with SVM in credit scoring,” Expert Systems with Applications, Vol. 37, No. 7, pp. 4902-4909, 2010.

[13] M. Y. Chen, “Using a hybrid evolution approach to forecast financial failures for Taiwan-listed companies,” Quantitative Finance, Vol. 14, No. 6, pp. 1047-1058, 2014.
[14] W. Chen, C. Ma, and L. Ma, “Mining the customer credit using hybrid support vector machine technique,” Expert Systems with Applications, Vol. 36, No. 4, pp. 7611-7616, 2009.

[15] L. C. Chi, and T. C. Tang, “Impact of reorganization announcements on distressed-stock returns,” Economic Modelling, Vol. 24, No. 5, pp. 749-767, 2007.

[16] E. Cholongitas, M. Senzolo, D. Patch, S. Shaw, C. Hui, and A. K. Burroughs, “Review article: scoring systems for assessing prognosis in critically ill adult cirrhotics,” Alimentary Pharmacology and Therapeutics, Vol. 24, No. 3, pp. 453-464, 2006.

[17] C. L. Chuang, and S. T. Huang, “A hybrid neural network approach for credit scoring,” Expert Systems, Vol. 28, No. 2, pp. 185-196, 2011.

[18] J. N. Crook, D. B. Edelman, and L. C. Thomas, “Recent developments in consumer credit risk assessment,” European Journal of Operational Research, Vol. 183, No. 3, pp. 1447-1465, 2007.

[19] V. S. Desai, J. N. Crook, and G. A. Jr. Overstreet, “A comparison of neural networks and linear scoring models in the credit union environment,” European Journal of Operational Research, Vol. 95, No. 1, pp. 24-37, 1996.

[20] G. Dong, K. K. Lai, and J. Yen, “Credit scorecard based on logistic regression with random coefficients,” Procedia Computer Science, Vol. 1, No. 1, pp. 2463-2468, 2010.

[21] M. A. H. Farquad, V. Ravi, Sriramjee, and G. Praveen, “Credit Scoring using PCA-SVM hybrid model,” Communications in Computer and Information Science, Vol. 142, No. 1, pp. 249-253, 2011.

[22] D. Feldman, and S. Gross, “Mortgage default: classification tree analysis,” The Journal of Real Estate Finance and Economics, Vol. 30, No. 4, pp. 369-396, 2005.

[23] S. Finlay, “The management of consumer credit: Theory and practice,” Basingstoke, UK: Palgrave Macmillan, 2008.

[24] M. Fioramanti, “Predicting sovereign debt crises using artificial neural networks: A comparative approach,” Journal of Financial Stability, Vol. 4, No. 2, pp. 149‐164, 2008.

[25] R. Geng, I. Bose, and X. Chen, “Prediction of financial distress: An empirical study of listed Chinese companies using data mining,” European Journal of Operational Research, Vol. 241, No. 1, pp. 236-247, 2015.

[26] A. Gepp, and K. Kumar, “Predicting Financial Distress: A Comparison of Survival Analysis and Decision Tree Techniques,” Procedia Computer Science, Vol. 54, No. 4, pp. 396‐404, 2015.

[27] A. Ghodselahi, “A hybrid support vector machine ensemble model for credit scoring,” International Journal of Computer Applications, Vol. 17, No. 5, pp. 1-5, 2011.

[28] J. Hadden, A. Tiwari, R. Roy, and D. Ruta, “Computer assisted customer churn management: state-of-the-art and future trends,” Computers and Operations Research, Vol. 34, No. 10, pp. 2902-2917, 2007.

[29] A. Z. Hamadani, T. Rezvan, and A. Moghadam, “An Integrated genetic-based model of Naïve Bayes networks for credit scoring,” International Journal of Artificial Intelligence & Applications, Vol. 4, No. 1, pp. 85-103, 2013.

[30] J. Holland, “Adaptation in natural and artificial systems,” The University of Michigan Press, Reissued by The MIT Press, 1975.

[31] H. Ince, and B. Aktan, “A comparison of data mining techniques for credit scoring in banking: A managerial perspective,” Journal of Business Economics and Management, Vol. 10, No. 3, pp. 233-240, 2009.

[32] F. J. L. Iturriaga, and I. P. Sanz, “Bankruptcy visualization and prediction using neural networks: A study of U.S. commercial banks,” Expert Systems with Applications, Vol. 42, No. 6, pp. 2857-2869, 2015.

[33] C. P. Jacovides, F. S. Tymvios, J. Boland, and M. Tsitouri, “Artificial Neural Network models for estimating daily solar global UV, PAR and broadband radiant fluxes in an eastern Mediterranean site,” Atmospheric Research, Vol. 152, No. 1, pp. 138-145, 2015.

[34] Y. Jiang, “Credit Scoring Model Based on the Decision Tree and the Simulated Annealing Algorithm,” Computer Science and Information Engineering, Vol. 4, pp. 18-22, 2009.
[35] H. Jo, I. Han, and H. Lee, “Bankruptcy prediction using case-based reasoning, neural networks, and discriminant analysis,” Expert Systems with Applications, Vol. 13, No. 2, pp. 97-108, 1997.

[36] D. N. Joanes, “Rejecting inference applied to logistic regression for credit scoring,” IMA Journal of Mathematics Applied in Business and Industry, Vol. 5, No. 4, pp. 35-43, 1993.

[37] A. Khashman, “Credit risk evaluation using neural networks: Emotional versus conventional models,” Applied Soft Computing, Vol. 11, No. 8, pp. 5477-5484, 2011.

[38] H. S. Kim, and S. Y. Sohn, “Support vector machines for default prediction of SMEs based on technology credit,” European Journal of Operational Research, Vol. 201, No. 3, pp. 838-846, 2010.

[39] K. Kim, and H. Ahn, “A corporate credit rating model using multi-class support vector machines with an ordinal pairwise partitioning approach,” Computers & Operations Research, Vol. 39, No. 8, pp. 1800-1811, 2012.

[40] H. C. Koh, W. C. Tan, and C. P. Goh, “A two-step method to construct credit scoring models with data mining techniques,” International Journal of Business and Information, Vol. 1, No. 1, pp. 96-118, 2006.

[41] E. Kolay, and T. Baser, “Estimating of the dry unit weight of compacted soils using general linear model and multi-layer perceptron neural networks,” Applied Soft Computing, Vol. 18, No. 1, pp. 223-231, 2014.

[42] S. Kotsiantis, B. D. Kanellopoulos, and P. Pintelas, “Data Preprocessing for Supervised Leaning,” International Journal of Computer Science, Vol. 1, No. 2, pp. 111-117, 2006.

[43] E. K. Laitinen, “Predicting a corporate credit analyst’s risk estimate by logistic and linear models,” International Review of Financial Analysis, Vol. 8, No. 2, pp. 97-121, 1999.

[44] T. H. Lee, and S. Jung, “Forecasting creditworthiness: Logistic vs. artificial neural net,” The Journal of Business Forecasting Methods and Systems, Vol. 18, No. 4, pp. 28-30, 2000.

[45] T. S. Lee, C. C. Chiu, Y. C. Chou, and C. J. Lu, “Mining the customer credit using classification and regression tree and multivariate adaptive regression splines,” Computational Statistics and Data Analysis, Vol. 50, No. 4, pp. 1113-1130, 2006.

[46] D. Liang, C. F. Tsai, and H. T. Wu, “The effect of feature selection on financial distress prediction,” Knowledge-Based Systems, Vol. 73, No. 1, pp. 289-297, 2015.

[47] Z. Lixin, and W Long, “Construction of the Logistic Regression Estimation Model in Early Warning on Pure Financial Indicators”, The Open Cybernetics & Systemics Journal, Vol. 9, No. 1, pp. 2055-2059, 2015.

[48] W. M. Makuch, “The basics of a better application score,” In E. Mays (Ed.), Handbook of credit scoring, Chicago: Glenlake Publishing Company, 2001.

[49] E. Mays, “Handbook of credit scoring,” Global Professional Publishing, pp. 23-56, 2001.

[50] J. Mylonakis, and G. Diacogiannis, “Evaluating the Likelihood of Using Linear Discriminant Analysis as A Commercial Bank Card Owners Credit Scoring Model,” International business research, Vol. 3, No. 2, pp. 9-20, 2010.

[51] G. Paleologo, A. Elisseeff, and G. Antonini, “Subagging for credit scoring models,” European Journal of Operational Research, Vol. 201, No. 2, pp. 490-499, 2010.

[52] M. Paliwal, and U. A. Kumar, “Neural networks and statistical techniques: A review of applications,” Expert Systems with Applications, Vol. 36, No. 1, pp. 2-17, 2009.

[53] S. Piramuthu, “Financial credit-risk evaluation with neural and neurofuzzy systems,” European Journal of Operational Research, Vol. 112, No. 2, pp. 310-321, 1999.

[54] R. Setiono, B. Baesens, and C. Mues, “Rule extraction from minimal neural networks for credit card screening,” International Journal of Neural Systems, Vol. 21, No. 4, pp. 265-276, 2011.

[55] A. Steenackers, and M. J. Goovaerts, “A credit scoring model for personal loans,” Insurance: Mathematics and Economics, Vol. 8, No. 1, pp. 31-34, 1989.

[56] K. Y. Tam, and M. Y. Kiang, “Managerial applications of the neural networks: The case of bank failure predictions,” Management Science, Vol. 38, No. 7, pp. 926-947, 1992.

[57] T. C. Tang, and L. C. Chi, “Neural networks analysis in business failure prediction of Chinese importers: A between-countries approach,” Expert Systems with Applications, Vol. 29, No. 2, pp. 244-255, 2005.

[58] N. Tripathi, and A. Jadeja, “Stochastically reducing overfitting in deep neural network using dropout,” International Journal of Innovative Science, Engineering & Technology, Vol. 2, No. 5, pp. 465-469, 2015.

[59] C. F. Tsai, and M. L. Chen, “Credit rating by hybrid machine learning techniques,” Applied soft computing, Vol. 10, No. 2, pp. 374-380, 2010.

[60] G. Wang, J. Ma, L. Huang, and K. Xu, “Two credit scoring models based on dual strategy ensemble trees,” Knowledge-Based Systems, Vol. 26, pp. 61-68, 2012.

[61] M. L. Wang, Z. Y. Feng, and Y. H. Chang, “A preliminary application of genetic algorithm on slope stability analysis,” Journal of Soil and Water Conservation, Vol. 2, pp. 65-82, 2010.

[62] D. West, “Neural network credit scoring models,” Computers and Operations Research, Vol. 27, pp. 1131-1152, 2000.

[63] B. W. Yap, S. H. Ong, and N. H. M. Husain, “Using data mining to improve assessment of credit worthiness via credit scoring models,” Expert Systems with Applications, Vol. 38, No. 10, pp. 13274-13283, 2011.

[64] H. Zhu, P. A. Beling, and G. A. Overstreet, “A study in the combination of two consumer credit scores,” Journal of the Operational Research Society, Vol. 52, No. 9, pp. 974-980, 2001.