Research On Applications Of Data Mining Technology In The Banking Credit Management

As the banking pays more and more attention to the information construction, information technology has been penetrated into various agencies, services and even all aspects of the banking. Most commercial banks in china have been established a complete set of information management system, which takes the accounting system as the core, and takes the personal credit management system, the credit management system, the intermediate business system and the management accounting system as a set of external support systems. As a result,more and more business data are collected in the database of the bank. However, many important decisions of the bank are not based on those data involved rich information, but on managers'intuition. Because the bank doesn't have the tool which can extract the useful information from mass business data, those mass data in the database turn to be"the data grave"and even to be the burden of the bank. The emergence of data mining technology can solve this problem.Data Mining is an interdisciplinary technology, combining technology and methods of database, machine learning, statistics and other fields. Data Mining is able to distill useful information and knowledge hidden in large quantities of data, which will greatly help managers to make more wise and correct decisions, guard against and manage credit risks.This thesis focuses on the research on risk management of credit management, carries out credit risk assessment by Data Mining technology. First, it introduces the technology and methods of data mining, analyzes the features of credit management in China's banks. Then we design a credit management system based on data mining. According to the industry standard of Data Mining technology CRISP-DM, the research sets up the credit risk assessment model with the six steps: business understanding, data understanding, data preparation, modeling, evaluation and deployment. After preprocessing on data, three methods for setting up models are utilized in this thesis: decision tree, neural net and logistic regression. And taking model result, mis-classification rate as general evaluation criteria, compare the three models, and conclude credit risk assessment criteria. Finally, this thesis focuses on summaries of the woks, discusses the limitation of this research, and the expectation next is mentioned in the end.