Applying Decision Tree Algorithm to construct a Financial Institution Inquire Bad Debt Account Taxation Information Model

• Main Authors: HSIEH,CHING-HUI, 謝菁惠
• Other Authors: Lan ,Tian-Syung
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/18436601087164883035

碩士 === 育達科技大學 === 資訊管理所 === 104 === Currently, the economic condition in Taiwan faces a downturn.The non-performing loans rations of financial institutions are too high. How to lower the ratios and liquidate the non-performing assets are the important issues for financial institutions. If those assets can be reclaimed through liquidating, it will be a good method to increase financial resources. In order to avoid wasting unnecessary operation expenses and realize the efficient resource management, we use Payable, Inquire date, Debt Obligation and etc. as requirements of refining those that should be looked into when we are inquiring bad accounts’ taxation information. In this study, the Decision Tree algorithm is used to construct an inquiring bad accounts’ taxation information model for financial institutions. It is based on, “The essential of acceptance for creditor inquiring about the taxation information for debtor”, released by Financial Information center of Ministry of Finance in 2013.06.19,importing those data into database. There are 5456 research samples, adding up association guarantors, ticket drawer, inheritors and etc. to a total amount of 9243 data. The total amount of book credit is NT$25.6 billion. We construct through the well-known ‘Decision Tree’ in the field of ‘Data Mining’. In the result, we discover that inquiring bad accounts’ taxation information for financial institutions can be categorized fast and efficiently through decision tree algorithm. With this method, financial institutions are able to select those samples that are suitable for inquiring and reach efficient management. It takes a total of NT$4,621,500 to check the full 9,243 pieces of taxation information. The data mining research method can be used to carry out decision tree model analysis to identify the 672 bad debts written off taxable information to be checked, saving a total cost of NT$4,285,500 if each check takes NT$336,000.