| As China's reform and openness develops deeply, the people's consuming conception renovated gradually, with booming of personal credit in the banking industry. Being the instrument of payment in daily consume and the non-target revolving loan, the business of credit card, since 2003,has attracted the attention of commercial banks at home and abroad following loans of motors and houses. This also attributes to commercial banks' aiming at its potential high profit and giant market space.As we all know, risk always accompany with return for most banking business. The business without risk is not all good, and escaping from the risk means loss of chances for earning. The success of credit cards business depends on risks management, that is, to realize the maximum of profit through risks undertaking and effective management.Since the business of credit cards possesses many lines of risks, such as credit risk, interest rate risk, exchange rate risk and liquidity risk etc., and the risks exist in every step of business, it is hard to establish a comprehensive and effective managing system for the risks of credit cards in one step. Though introducing the advanced experience from foreign banks, we need a long time to study and digest for the reason of the different national situations.So this thesis focuses on the research on credit scoring of those applying for credit cards, which belongs to the crediting link within the risk management of credit cards, and develops a managing system based on the studying result.Credit scoring is a quantitative assessment for credit cards applicants. It has two basic hypotheses: the first is that, for a human being, the past activities can be regarded as tokens of his future activities; the second is that people with similar backgrounds and behavior c will have the same activities. Both are supported by the great database of credit agencies in U.S. and its statistical analysis wholly.The methods of credit rating can be divided into three classifications in nature: rules based scoring, behavior based scoring and neural network modeling. The data mining method can be classified into behavior based scoring.Data mining is the process to collect the hidden, unknown but potentially useful information and knowledge from numerous, incomplete, noisy, obscure and random data, which aggregates database, artificial intelligence, statistics, invisibility and parallel calculation as a cross subject. The most attractive point is the ability to establish a model forecast, not merely to review. Data mining has been applied to all over the business, operation and decision-making as a key to success in competition in U.S.By means of data mining, the establishment of credit scoring on applicants of credit cards can provide an objective, accurate and coincident standard for managing the credit risk in banking industry. To set up a mathematic model and to construct the first defense line, a bank can define the characteristics of its good and bad clients' backgrounds, such as ages, incomes, professions and educational levels from a number of data about clients' backgrounds, behaviors and credit, and measure the contributing weight form various characters. The tool for mining is SAS systemic software, mainly the module Enterprise Miner. According to the methodology of SEMMA provided by SAS Co., the research sets up the credit scoring model with the five steps: sample, explore, modify, model and assess. Two methods for setting up models are utilized in this thesis: one is decision tree, and the other is logistic regression.This study is to establish the model of decision tree and logistic regression and make comparative analysis by use of exploring the basic and pertinent information of piles of clients in order to seek optimum. On the basis of optimum model companies are able to analyze and score status of credit risk, then to attain the optimum scoring value to distinguish good clients from bad ones, and to make quick responses to lots of applicants, fur... |
PDF Full Text Request