Using Big Data Analysis to Reduce Fiduciary Loan Delinquency Rate

• Main Authors: Yen-Hui Yao, 姚堰輝
• Other Authors: Chien-Yi Huang
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/7vt3rx

碩士 === 國立臺北科技大學 === 管理學院工業工程與管理EMBA專班 === 106 === Credit card, wealth management, consumer loan are main financial products of banking business. Consumer loan is an unsecured credit loan features high lending rate, high return rate and high efficiency, which stands as one of major profitable products among banking consumer financial products. Delinquent risk analysis and management under the factor of unsecured background is an important subject for bank. So how can bank minimize the loss of delinquency by using effective credit scoring model. The aim of this study is starting from selecting the credit card holders of a bank as objects, following with selecting part of those credit card holders as base who has gained the approval and obtained credit loan from a branch office of this bank within a designed period of time, and random sampling 2 groups from the base, which classified by payment records include normal payment and overdue. Next step is using Discriminant analysis of SPSS 19.0 statistical software to analyze significant factors that may cause overdue among these sampling. The result of analysis indicates that annual income, seniority, job position, etc. are the significant factors which affect default. After these important factors that may affect default are found, an evaluation pattern of default risk for unsecured loan can be built and used for reference of risk management of unsecured loan accordingly. Following with factor analysis and principal component analysis for choosing and integrating these important factors, a pattern of default prediction can be built by using classification and regression tree. The overall prediction accuracy of this pattern is up to 79%.