Customer's Marketing Response Module Design Based On SAS

Credit card is a popular instrument of payment and consumer credit internationally and is considered the electronic money of the modern society. Being safe, convenient and swift, it boasts powerful vitality and is the symbol of the modernization of financial industries.China is the most promising credit card market globally. Taking off relatively late, it's a long way behind developed countries in theory and application. However, having realized the importance of credit card as an essential part of the payment system, commercial banks across China began expanding their credit card businesses since 2003, the first year of credit cards, driving up the market share of credit card payment. Over the past two years, with the growing acceptance by the general public and the ever-improving banking card service environment, credit card payment has become a key index in boosting consumer spending, promoting the development of the retail market of consumer goods and stimulating domestic demand.However, the increasing number of credit cards issued by different banks has a tendency of homogenization and the credit card business enters a development bottleneck with a slow growth rate. How to gain advantage in the fierce competitive and how to select suitable target customers before credit card product marketing becomes a problem which has been concern by many commercial banks. So, data-mining has become an important technology in creating marketing response module, exploring laws and patterns in data, providing solutions in credit card marketing management.This dissertation is going to study the development of data-mining technology, analyze the commercial bank data management problems, design a marketing response module according to a marketing case and at last calculate the probability of response. Early in the module designing, it takes a lot of time in data processing and erroneous data removing in order not to affect module's predictive capability. Then we use logistic regression to design module and assess its performance according to the training data and validation data. The little differences compared by two LIFT-tables shows that module's performance is satisfying and at the end of the dissertation, details has been shown in the module processing.