An Application of Anomaly Detection to P2P Lending

• Main Authors: LIN, PO-JUI, 林柏叡
• Other Authors: KAO, MING-SUNG
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/8v7e7g

碩士 === 輔仁大學 === 金融與國際企業學系金融碩士班 === 107 === Peer-to-peer (P2P) lending has developed rapidly. More and more people use this method to borrow money. and before lending to people how to control the risk and after lending how to monitor are very important. For the platform, it can be monitored during the lending process, and it can be detected in advance before it is completely breached. Thus this study uses abnormal detection analyzes the US P2P platform on how to control management and monitoring.In the past, monitoring was often performed separately for individual variables, without considering the influence between variables. In the past, monitoring was often carried out separately for individual variables, and did not take into account the influence between variables. This study used five important variables to divide the interval, and found out the data of high default and low default. Data status, this study did a multivariate anomaly detection method, and considering the importance of the order of observation between variables, the results of this study found that the selected priority variables will vary with the situation. Good borrowers prioritize the use of all bank balances, while dangerous borrowers prioritize the monitoring of debt-to-income ratios and then match the remaining four variables to achieve more complete testing than ever before.