Research On Credit Card Customer Segmentation Model Based On Associative Classification

Along with more and more competitive credit card market and diversification of client, most of the local bank has stepped into the international operation pattern of "Client-first". Client management is being the focus for competing and developing of the bank. Banks have to take the initiative to conduct customer segmentation in order to maintain leadership position and ever-increasing its value. Most of the traditional customer segmentation based on experience or simple statistical methods can not meet the growing volume of data, as well as the complex analysis of the business. The classification based on data digging methods appear to be as a new customer segmentation solution under mass of data, but also as an effective tool for targeted marketing.Aim at the exits large data of credit card data in the bank, this paper which based on the popular and international research use an associative classification algorithm to segment the credit card customers to enhance customer satisfaction and their own the value of competition by achieving the marketing of targeted customers. Main researches are as follows:Firstly, author does the research on association classification deeply, especially on the classic algorithm, which builds the theories precondition for the improved association classification algorithm put forward in this article.Secondly, construct the index system of credit card customers' subdivision. By studying the behavior of credit cards using, and taking into account the implied value at the same time, this text suggests to conduct a comprehensive analysis from the customer's personal characteristics, the customer's consuming behavior and customer value analysis. On that basis, construct for the index system of credit card customers, and as an evidence for customer segmentation.Thirdly, propose a correlation-based association and classification algorithm ACBC. Above all, using of CM with a combination of greedy algorithm, and considering CM as a rule of quality assessment criteria, it'll direct deletes irrelevant or weakly relevant rules when rules creation stage. Finally, test data sets and forecasts by using the classifier. As proved by experiments, ACBC algorithm has resulted as better classification ability, and also reducing computation time and storage space occupancy. Fourthly, a credit card customer segmentation model is put forward in the article, which is based on association classification algorithm. Using ACBCM as the critical technology and using bank customer information as data source to mine the associative rules between the frequent property sets and class label. This model realize the function of credit card customer segmentation, the classification results can build the foundation for personalized marketing.