A Study On Mining Changes Of Customer Behavior Based On Association Rule And Its Application

Currently, two transformations have happened in Chinese market: one is from seller's market to buyer's market and the other is from the scarcity of products to the scarcity of customers. Competitions, especially on customer resource, among enterprises are much fiercer. They become convinced that customer is the most important resource. On the other hand, with the rapid development of computer and information technology, more enterprises make use of them to strengthen their management, such as financial system, Enterprise Resource Planning (ERP) and Customer Relation Management (CRM).When information systems are used for a long time, decision-makers begin to conceive an idea whether they can extract useful information from large-scale databases to help them summarize and even predict the changes of customers' demands and purchasing behavior. The emergence and development of Data Mining (DM) technology make it possible. Association Rule (AR), one of the most active branches of DM, has been used in the analysis on customer's behavior.The core of this paper is to analyze the trends of customers' behavior dynamically using AR, to measure the trends with one specific measurement theory and finally to demonstrate them to decision-makers with the technology of visualization. This paper firstly gives a systematical introduction on CRM and DM, especially on AR and its algorithms. An improved Apriori algorithm, based on shrinking dataset, is introduced intensively and proved to be more effective than the traditional Apriori algorithm in one test. Then, it introduces the measurement theory of changing of customer behavior: mathematical definitions of changing patterns, the method and steps to categorize patterns. In the core chapter, it shows the whole procedure of 4-step technical routine: building data warehousing, data ETL, mining association rule-sets of two periods and measurement of changing customer's behavior. It also gives details of what the system does in each step. At last, this paper summarizes different kinds of patterns generated in the system, explains the meaning of each kind of pattern through instances, and gives marketing suggestions for each specific pattern to the marketing decision-makers of that enterprise.