臺灣博碩士論文加值系統

P2P 借貸是近年國外金融科技(Fintech)中極為熱門的業務。基本上，P2P 借 貸業者透過平台媒合貸款人及債權人，並提供風險評估。本研究透過分析 Kaggle 網站中數據專家針對 Lending Club 資料集所提供之資料分析方法，進行方法的比 較與探討，並思考國內商業銀行如何利用相關技術來拓展 P2P 借貸之相關業務。 研究發現利用機器學習方法可有效地找出貸款人違約的關鍵因子，並進而有效預 測貸款人違約機率。在商業銀行業務拓展的思考下，可運用 P2P 借貸技術對商業 銀行高風險客戶提供資金貸款服務。

In recent years, P2P (Peer to Peer Lending) has become one the of most popular financial technology (FinTech) business worldwide. P2P lending company is the media that connects the investors and borrowers. They also provide the risk analysis for their customers. This study is to exam the analytic technology from Kaggle - a data science and machine learning website. The statistic experts and data scientists at Kaggle analyzed and cross-exam Lending Club, a typical P2P lending company, that leads the online lending platform. Kaggle’s analysis helps us understand how we can potentially utilize the relevant P2P data analytic technology for Taiwan’s business banks. The study found that the use of machine learning method can effectively identify the key factors of the borrower’s default, and thus effectively predict the probability of lender's risk. Therefore, P2P lending technology can be used to determine the loan services to the high-risk clients in the development of commercial banks.

摘要 i
ABSTRACT ii
誌謝 iii
圖目錄 vi
表目錄 vii
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 3
1.3 研究流程與論文架構 4
第二章 文獻回顧 5
2.1 傳統借貸與金融創新 P2P 借貸模式 5
2.1.1 傳統借貸 5
2.1.2 P2P 借貸 6
2.1.3 傳統借貸與金融創新 P2P 借貸比較 7
2.2 數據分析與機器學習分類方法 8
2.2.1 資料視覺化(Data Visualization) 8 2.2.2 決策樹(Decision Tree) 9
2.2.3 隨機森林(Random Forest) 10
2.2.4 混淆矩陣(Confusion Matrix) 10
第三章 研究方法與討論 12
3.1 資料視覺化分析 Lending Club 資料集 12
3.1.1 清償貸款時間分析 12
3.1.2 借貸目的分析 13
3.1.3 借貸目的與清償還款天數的平均值 13
3.1.4 貸款利息的分布與數量 14
3.2 機器學習方法預測整理 15
3.2.1 A 專家:利用隨機森林(Random Forest)預測還款違約 16
3.2.2 B 專家:利用決策樹(Decision tree)預測借貸目的為信用卡(credit card)、醫療(medical)以及債務(debt)的還款違約 19
3.3 研究方法結論 26
第四章 研究結論與建議 30
4.1 研究結論 30
4.2 研究建議 30
參考文獻 32

1. Asuru, C. (2017). "Visualizations." Retrieved 07 02, 2017, from https://www.kaggle.com/casuru/visualizations.
2. Chatterjee, S. K., S. Das, K. Maharatna, E. Masi, L. Santopolo, I. Colzi, S. Mancuso and A. Vitaletti (2017). "Comparison of decision tree based classification strategies to detect external chemical stimuli from raw and filtered plant electrical response." Sensors and Actuators B-Chemical 249: 278-295.
3. Chishti, S. and J. Barberis (2016). FinTech 金融科技聖經:全球 86 位 FinTech 先 驅，第一手公開「金融科技做什麼?怎麼做?」 (唐祖蔭、劉奕吟、陳世杰， 譯). 台灣, 商業周刊.
4. Google. (2017). "Google Trends." Retrieved 06 17, 2017, from https://trends.google.com.tw/trends/explore?geo=TW&q=fintech.
5. gratipine. (2017). "Exploration of purposes, time and interest rates | Kaggle." Retrieved 07 02, 2017, from https://www.kaggle.com/gratipine/exploration-of- purposes-time-and-interest-rates.
6. JakubKúdela. (2017). "Analysis + Decision Tree Default Prediction | Kaggle." Retrieved 07 02, 2017, from https://www.kaggle.com/kubo56/analysis-decision-tree- default-prediction.
7. Josh. (2017). "Predicting Loan Status with Python | Kaggle." Retrieved 07 02, 2017, from https://www.kaggle.com/jlrsource/predicting-loan-status-with-python.
8. Kaggle. (2017, 2015-10-12 ). "Home | Kaggle." Retrieved 06 27, 2017, from https://www.kaggle.com/wiki/Home.
9. Kan, W. (2017). "Lending Club Loan Data | Kaggle." Retrieved 07 06, 2017, from https://www.kaggle.com/wendykan/lending-club-loan-data.
10. Lee, S. (2017). "Evaluation of Mobile Application in User’s Perspective: Case of P2P Lending Apps in FinTech Industry." KSII Transactions on Internet and Information Systems 11(2): 1105-1117.
11. LendingClub. (2017). "About Us | LendingClub." Retrieved 06 28, 2017, from https://www.lendingclub.com/public/about-us.action.
12. Lewis, H. and M. Brown (2001). "A generalized confusion matrix for assessing area estimates from remotely sensed data." International Journal of Remote Sensing 22(16): 3223-3235.
13. Osterwalder, A. and Y. Pigneur (2010). 獲利世代:自己動手，畫出你的商業模式 (尤傳莉，譯). 台灣, 早安財經.
14. Zetzsche, D. A., R. P. Buckley, D. W. Arner and J. N. Barberis (2017). "From FinTech to TechFin: The Regulatory Challenges of Data-Driven Finance." European Banking Institute Working Paper Series 2017 6: 5.
15. Zopa. (2017). "About Zopa." Retrieved 06 28, 2017, from https://www.zopa.com/about.
16. 王蓬博. (2017, 2017-05-17). "易观:2017 年第 1 季度中国第三方支付移动支付市 场交易规模 188091 亿元人民币." Retrieved 06 27, 2017, from https://www.analysys.cn/analysis/22/details?articleId=1000762.
17. 吴仍康 (2016). "基于机器学习方法的上证综合指数的预测分析." Hans Journal of Data Mining 6: 5.
18. 拍拍贷. (2017). "信息披露_拍拍贷公司简介_拍拍贷." Retrieved 06 28, 2017, from http://www.ppdai.com/help/xinxipilu.
19. 蚂蚁金融服务集团. (2017). "蚂蚁家族." Retrieved 06 27, 2017, from https://www.antfin.com/family.htm.
20. 張瑞方 (2015). 影響消費者對銀行業信用貸款的使用意願與否之研究-以台南地區的信用貸款為例. 碩士, 國立成功大學.
21. 梅驊 (2016). 全球 P2P 網路借貸現況與台灣發展趨勢探討. 碩士, 國立臺灣科技大學.
22. 維基百科. (2017, 20170608). "典當業 - 維基百科，自由的百科全書."
Retrieved 06 27, 2017, from https://zh.wikipedia.org/wiki/典當業.
23. 維基百科. (2017, 2016 12 17). "資料視覺化 - 維基百科，自由的百科全書."
Retrieved 06 29, 2017, from https://zh.wikipedia.org/wiki/數據可視化.