跳到主要內容

臺灣博碩士論文加值系統

(44.200.171.156) 您好!臺灣時間:2023/03/22 03:13
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:林亮妤
研究生(外文):LIN, LIANG-YU
論文名稱:深度學習自編碼演算法在信用風險之應用
論文名稱(外文):The Applications of Deep Learning Autoencoder Approach on Credit Risk
指導教授:張揖平張揖平引用關係
指導教授(外文):CHANG, YI-PING
口試委員:李沃牆吳錦全
口試委員(外文):LEE, WO-CHIANGWU, CHIU-CHUAN
口試日期:2019-06-08
學位類別:碩士
校院名稱:東吳大學
系所名稱:財務工程與精算數學系
學門:數學及統計學門
學類:其他數學及統計學類
論文種類:學術論文
論文出版年:2019
畢業學年度:107
語文別:中文
論文頁數:65
中文關鍵詞:深度學習類神經網路機器學習自編碼信用風險
外文關鍵詞:Deep LearningArtificial Neural NetworksMachine LearningAutoencoderCredit Risk
相關次數:
  • 被引用被引用:0
  • 點閱點閱:241
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:0
機器學習(Machine Learning)是近年興起的議題,其中深度學習(Deep Learning)為機器學習的一個分支,透過深度神經網路讓電腦自行分析資料找出特徵值(Feature),進一步針對問題作出合適的判斷。隨著科技的發展,信用評分模型、股票市場預測等金融議題也加入深度學習的分析方法,使得深度學習在金融資料上的應用成為未來創新的發展。本研究使用 Kaggle 網站上的信用卡交易紀錄為研究對象,以 2006 年 Geoffrey and Salakhutdinov 提出的自編碼(Autoencoder)進行模型建置及詐欺偵測,並加入美國 P2P 貸款資料進行比較。研究結果顯示,自動編碼器在信用卡交易紀錄資料上可以有效地偵測到該筆交易是否具有詐欺的效果,但是在美國 P2P 貸款資料卻無法有效的偵測出是否具有違約的特徵,因此自動編碼器是否適用於所有類型的信用風險議題仍具爭議。
Machine Learning is a topic that has emerged these years. Deep Learning is a branch of machine learning. Through deep neural networks, computers can analyze data by themselves to find features and further adapt to the problem. With the development of science and technology, financial issues such as credit rating model and stock market forecasting have also been added to the analysis method of deep learning, making the application of deep learning in financial data become the development of future innovation. This study used the credit card transaction on the Kaggle website as the subjects, using the Autoencoder proposed by Geoffrey and Salakhutdinov in 2006 for training an Autoencoder Neural Network in unsupervised fashion for anomaly detection in credit card transaction data and compare to the data from a peer to peer lending company based in the United States, trying to predict if a loan will default or not. The research results show that the Autoencoder can effectively detect whether the transaction has the effect of fraud on the credit card transaction data, but the peer to peer lending data cannot effectively detect whether it has the characteristics of default, so the Autoencoder not applicable to all types of credit risk issues.
1. 緒論 1
2. 文獻回顧 4
3. 研究方法 8
3.1. 模型函數 11
(1) 模型層數(Layer) 11
(2) 激活函數(Activation Function) 11
(3) 損失函數(Loss Function) 12
(4) 優化器(Optimizer) 13
3.2. 模型指標 15
(1) Confusion Matrix 15
(2) 接收者操作特徵曲線(Receiver Operating Characteristic curve, ROC curve) 15
(3) AUC(Area Under the Curve of ROC) 16
4. 實證分析 17
4.1 信用卡詐欺資料 17
4.2 美國地區的 P2P 貸款資料 28
4.3 結果比較 40
5. 結論與未來研究方向 43
參考文獻 44
附錄 46
附錄1 46
附錄2 50
附錄3 51
附錄4 53
附錄5 54
附錄6 58
附錄7 62


Hinton, G. E. and Salakhutdinov, R. R. (2006). Reducing the Dimensionality of Data with Neural Networks. Science, Volume 313, Issue 5786, pp. 504-507.
Janio, M. B. (2018). Risk analysis and metrics. Retrieved from https://www.kaggle.com/janiobachmann/lending-club-risk-analysis-and-metrics.
Kingma, D. P. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980. https://arxiv.org/pdf/1412.6980.pdf.
Mandrekar, J. N. (2010). Receiver operating characteristic curve in diagnostic test assessment. Journal of Thoracic Oncology Volume 5, Issue 9, pp. 1315-1316.
Lin, H. W., Termark, M. and Rolnick, D. (2017). Why does deep and cheap learning work so well? arXiv preprint arXiv: 1608.08225. https://arxiv.org/pdf/1608.08225.pdf.
Loshchilov, I. and Hutter. F. (2017). Fixing weight decay regularization in Adam. arXiv preprint arXiv: 1711.05101. https://arxiv.org/abs/1711.05101.
Official website of Kaggle. (2010). Retrieved from https://www.kaggle.com/.
Official website of Keras. (2015). Retrieved from https://keras.io/zh/.
Sashank J. R., Satyen, K. and Sanjiv, K. (2018). On the convergence of Adam and beyond. arXiv preprint arXiv:1904.09237. https://arxiv.org/pdf/1904.09237.pdf.
Svante, W., Kim, E. and Paul, G. (1987). Principal component analysis. Chemometrics and Intelligent Laboratory Systems Volume 2, Issues 1–3, pp. 37-52.
Venelin, V. (2017). Credit card fraud detection using Autoencoders in Keras — TensorFlow for Hackers. Retrieved from https://medium.com/@curiousily/credit-card-fraud-detection-using-autoencoders-in-keras-tensorflow-for-hackers-part-vii-20e0c85301bd.
Wang, W. (2017). Credit card fraud detection 1 - using Autoencoders in  TensorFlow. Retrieved from https://weiminwang.blog/2017/06/23/credit-card-fraud-detection-using-auto-encoder-in-tensorflow-2/.
Yann, L., Yoshua, B.and Geoffrey, H. (2015). Deep learning. Nature volume521 pp. 436–444.







電子全文 電子全文(網際網路公開日期:20240630)
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊