臺灣博碩士論文加值系統

由於2008年發生了金融海嘯，導致全球各地產生了嚴重的經濟危機，金融機構因此開始關注在信用風險分析上，因為若是借貸給沒有能力償還的申請人，金融機構將會損失慘重。而近年來信用風險分析的研究大多是把資料視為類別平衡的資料，但現實生活中沒有能力償還的申請人屬於少數類別，若是以一般分類方法處理可能會造成少數類別的預測正確率低落，而目前信用風險分析的研究尚無一組有效的特徵，因此本研究將使用加權法隨機森林以解決信用風險分析上資料不平衡的問題，除了可以改善少數類別低正確率的問題，也能找出信用風險資料中的有效特徵。本研究使用了四個金融相關的資料集，而研究結果也指出加權法隨機森林優於一般的隨機森林，最後會使用袋外錯誤率來找出在信用風險分析上有效且有意義的特徵。

Due to the financial crisis that occurred in 2008, which caused serious economic crises around the world, financial institutions began to pay attention to credit risk analysis. Financial institutions will suffer heavy losses if they lend to applicants who are unable to repay. In recent years, most of the research on credit risk analysis regards financial data are class-balanced. However, only few of applicants are unable to repay. If financial data are processed by general classification methods, the prediction accuracy on the minority class may be low. In addition, filtering effective features for credit risk analysis is critical. Therefore, this study will use weighted random forest instead of random forest to process imbalanced financial data such that not only the prediction accuracy on the minority class can be improved, but also the critical attributes for credit evaluation can be discovered. The experimental results obtained from four financial data sets show that weighted random forest outperforms random forest. The attributes found by applying out-of-bag error can provide meaningful insights in credit risk analysis.

目錄
摘要 I
誌謝 V
目錄 VI
表目錄 VIII
圖目錄 IX
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 2
1.3 論文流程與架構 3
第二章 文獻回顧 4
2.1 特徵選取 4
2.2 信用風險評估的分類方法 6
2.2.1 類神經網路 7
2.2.2 支援向量機 9
2.2.3 隨機森林 10
2.3 衡量測度 11
2.4 小結 14
第三章 研究方法 16
3.1 研究流程 16
3.2 資料前置處理 17
3.3 隨機森林 19
3.3.1 套袋法 20
3.3.2 決策樹 21
3.3.3 加權法隨機森林 22
3.3.4 利用隨機森林進行特徵重要性評估 24
3.4 衡量指標 25


第四章 實證研究 26
4.1 資料集介紹 26
4.2 加權法隨機森林重要參數介紹 27
4.3 加權法隨機森林實證研究 28
4.3.1 Give me some credit資料集之比較 29
4.3.2 Credit card scoring資料集之比較 31
4.3.3 Credit card fraud資料集之比較 33
4.3.4 German credit dataset資料集之比較 35
4.4 特徵重要性評估 37
4.5 小結 39
第五章 結論與建議 41
5.1 結論 41
5.1 建議 42
參考文獻 44
附錄一Give me some credit資料集特徵介紹 49
附錄二Credit card scoring資料集特徵介紹 50
附錄三Credit card fraud資料集特徵介紹 51
附錄四German credit dataset資料集特徵介紹 52

 
表目錄
表2.1、混淆矩陣表 11
表4.1、資料集介紹表 26
表4.2、Give me some credit實驗結果彙總表 29
表4.3、Credit card scoring實驗結果彙總表 31
表4.4、Credit card fraud實驗結果彙總表 33
表4.5、German credit dataset實驗結果彙總表 35
表4.6、Give me some credit 特徵重要性排序表 37
表4.7、Credit card scoring特徵重要性排序表 38
表4.8、German credit dataset特徵重要性排序表 39


 
圖目錄
圖2.1、wrapper特徵選取流程 5
圖2.2、filter特徵選取流程 5
圖2.3、類神經網路之神經元的構造圖 7
圖2.4、ROC曲線圖 13
圖3.1、研究方法步驟流程圖 17
圖3.2、隨機森林概念圖 20
圖4.1、Give me some credit 彙總圖 30
圖4.2、Give me some credit AUC比較圖 30
圖4.3、Credit card scoring 彙總圖 32
圖4.4、Credit card scoring AUC比較圖 32
圖4.5、Credit card fraud 彙總圖 34
圖4.6、Credit card fraud AUC比較圖 34
圖4.7、German credit dataset 彙總圖 36
圖4.8、German credit dataset AUC比較圖 36

陳錦村、許通安、林蔓蓁，(1996)。銀行授信客戶違約風險之預測。管理科學學報，13(1)，173-95。

葉怡成，(2003)。類神經網路模式應用與實作。台北：儒林圖書有限公司。

AghaeiRad, A., Chen, N., ＆ Ribeiro, B. (2017). Improve credit scoring using
transfer of learned knowledge from self-organizing map. Neural
Computing and Applications, 28(6), 1329-1342.

Aha, D. W., ＆ Bankert, R. L. (1996). A comparative evaluation of sequential
feature selection algorithms. In D. Fisher＆ H. J Lenz (Eds.), Learning from Data (pp. 199-206). New York:Springer.

Andric, K., ＆Kalpic, D. (2016). The effect of class distribution on classification algorithms in credit risk assessment. Proceedings of IEEE International Convention on Information and Communication Technology, Electronics and Microelectronics (MIPRO) 39, pp. 1241-1247..doi: 10.1109

Baesens, B., Van Gestel, T., Viaene, S., Stepanova, M., Suykens, J., ＆Vanthienen, J. (2003). Benchmarking state-of-the-art classification algorithms for credit scoring. Journal of the Operational Research Society, 54(6), 627-635.



Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5-32.
Chen, C., Liaw, A., ＆Breiman, L. (2004). Using random forest to learn imbalanced data. Doctoral dissertation, University of California, Berkeley.
Cortes, C., ＆Vapnik, V. (1995). Support-vector networks. Machine Learning, 20(3), 273-297.

Dahiya,S., Handa,S.S.,＆Singh,N.P. (2016). A rank aggregation algorithm for ensemble of multiple feature selection techniques in credit risk evaluation. International Journal of Advanced Research in Artificial Intelligence, 5(9), 1-8.

Danenas, P., ＆Garsva, G. (2015). Selection of support vector machines based classifiers for credit risk domain. Expert Systems with Applications, 42(6), 3194-3204.

Danenas, P., Garsva, G., ＆ Gudas, S. (2011). Credit risk evaluation model development using support vector based classifiers. Procedia Computer Science, 4, 1699-1707

Dash, M., ＆ Liu, H. (1997). Feature selection for classification. Intelligent Data Analysis, 1(3), 131-156.

Fawcett, T. (2006). An introduction to ROC analysis. Pattern Recognition Letters, 27(8), 861-874.

Freeman, J.A.＆Skapura, D. M.(1992). Neural networks: algorithms,applications and programming techniques. Massachusetts: Addison-Wesley.
Ghatasheh, N. (2014). Business analytics using random forest trees for credit risk prediction: A comparison study. International Journal of Advanced Science and Technology, 72, 19-30.

Gupta, D. K., ＆ Goyal, S. (2018). Credit risk prediction using artificial neural network algorithm. International Journal of Modern Education and Computer Science, 10(5), 9-16.

Guyon, I., ＆Elisseeff, A. (2003). An introduction to variable and feature selection. Journal of Machine Learning Research, 3, 1157-1182.

Huang, C. L., Chen, M. C., ＆ Wang, C. J. (2007). Credit scoring with a data mining approach based on support vector machines. Expert Systems with Applications, 33(4), 847-856.

Kohavi, R., ＆ John, G. H. (1997). Wrappers for feature subset selection. Artificial Intelligence, 97(1), 273-324.

Koutanaei, F. N., Sajedi, H., ＆Khanbabaei, M. (2015). A hybrid data mining model of feature selection algorithms and ensemble learning classifiers for credit scoring. Journal of Retailing and Consumer Services, 27, 11–23

Lessmann, S., Baesens, B., Seow, H. V., ＆ Thomas, L. C. (2015). Benchmarking state-of-the-art classification algorithms for credit scoring: An update of research. European Journal of Operational Research, 247(1), 124-136.

Li H., Huang C. N., Gao J., ＆ Fan X. (2005). The use of SVM for Chinese new word identification. InSu,K. Y., Tsujii, J., Lee, J. H., ＆Kwong O.Y. (Eds), Lecture Notes in Computer Science: vol. 3248.Natural Language Processing (pp.723–732). Berlin, Germany: Springer. doi: 10.1007/978-3-540-30211-7_76

Liang, D., Tsai, C. F., ＆ Wu, H. T. (2015). The effect of feature selection on financial distress prediction. Knowledge-Based Systems, 73, 289-297.

Liu, H., ＆Motoda, H. (1998). Feature extraction, construction and selection: A data mining perspective. New York:Springer.

Malekipirbazari, M., ＆Aksakalli, V. (2015). Risk assessment in social lending via random forests. Expert Systems with Applications, 42(10), 4621-4631.

Mandrekar, J. N. (2010). Receiver operating characteristic curve in diagnostic test assessment. Journal of Thoracic Oncology, 5(9), 1315-1316.

Mohammadi, N., ＆Zangeneh, M. (2016). Customer credit risk assessment using artificial neural networks. International Journal of Information Technology and Computer Science, 8(3), 58-66.
Pacelli,V.＆Azzollini,M. (2011). An artificial neural network approach for credit risk management. Journal of Intelligent Learning Systems and Applications,3, 103-112

Quinlan, J. R. (1986). Induction of decision trees. Machine Learning, 1(1), 81-106.

Rumelhart, D. E., Hinton, G. E., ＆ Williams, R. J. (1986). Learning
representations by back-propagating errors. Nature, 323(6088), 533-536.

Vinh, N. X., Zhou, S., Chan, J., ＆ Bailey, J. (2016). Can high-order dependencies improve mutual information based feature selection?.Pattern Recognition, 53, 46-58.

Yan, K., ＆ Zhang, D. (2015). Feature selection and analysis on correlated gas sensor data with recursive feature elimination. Sensors and Actuators B: Chemical, 212, 353-363.

Yao, J., ＆ Lian, C. (2016). A new ensemble model based support vector machine for credit assessing. International Journal of Grid and Distributed Computing, 9(6), 159-168.

Yu, L., ＆ Liu, H. (2004). Efficient feature selection via analysis of relevance and redundancy. Journal of Machine Learning Research, 5, 1205-1224.

Yu, L., Yang, Z., ＆ Tang, L. (2016). A novel multistage deep belief network based extreme learning machine ensemble learning paradigm for credit risk assessment. Flexible Services and Manufacturing Journal, 28(4), 576-592.

Zhang, Q., Wang, J., Lu, A., Wang, S., ＆ Ma, J. (2018). An improved SMO algorithm for financial credit risk assessment–Evidence from China’s banking. Neurocomputing, 272, 314-325.