Research On Customer Segmentation Model Based On Distributed Associative Classification

Propelled by the evolution of management theories, the demand of management practices and the advancement of information technology, Customer Relationship Management (CRM) has become one of the hottest topics for all enterprises around the world in the 21^st century. One of the fundamental functions in CRM is customer segmentation, which plays the pivotal role of supporting the key business process of customer acquisition, customer retention and customer value extension. The outcomes of customer segmentation have profound effects throughout the CRM process.Currently, large chain retail enterprises expand rapidly and develop into international operations with the adoption of networks and information technology. Commodity business has evoluted into "information business". A large number of chain stores, distribution centers and headquarters have constructed commercial data-sharing environment through network interconnection. In which, each node has collected a lot of transaction data on customer and creates the large-scale distributed data gold mine. Consequently, the early methods for customer segmentation can hardly meet the challenges of today's exploding data size and complex customer analysis. The emergence of new classification methods based on distributed data mining technologies has opened up new ways of in-depth customer segmentation.Based on the existing research, this paper builds a customer segmentation model named DCSM. The outline contents in this paper describes as follows:Firstly, we study the traditional theory on customer segmentation and the classical associative classification algorithms, including CBA and CMAR algorithms in order to use them in the customer segmentation tasks.Secondly, considering the weakness of customer segmentation method in chain retail enterprises, the distributed customer segmentation model DCSM (Customer Segmentation Model based on Distributed Associative Classification) is proposed, which takes the customer data in various distributed nodes as the data source, mobile agent operation platform Bee-gent system as the framework. It uses FDAC algorithm (Distributed Associative Classification Algorithm based on improved FP-tree) as the key technology to get global associative classifier, and use this classifier to realize the customer segmentation.Thirdly, the FDAC algorithm is proposed. It first generates local improved FP-tree at every station, and then transmits conditional FP-tree to construct global conditional FP-tree. Then, it introduces the concept of signification when mining global conditional FP-tree to generate initial global significant classification rules. At last, the rules are pruned with several strategies to construct an associative classifier. The algorithm uses far less communication overhead and improves efficiency of mining global classification rules. It also can guarantee the statistical significance of the discovered knowledge and increase the ability to discover implicit rules.Fourthly, based on the work outlined above, a data mining prototype system is designed and implemented for customer segmentation on chain retail enterprise. The main functions of this prototype system include data retrieval, data pre-processing, modeling and model application. Finally, supported by this prototype system, a specific business application is used. Through mining the personal characteristics, customer consumption behavior and customer value to get a customer segmentation result which improves the capability of business analysis, decision support and commodity management.