The Application Research Of Data Mining In SMS

Base on the digital TV industry background, this paper researches the data mining technology using in Subscriber Management System. The emphasis is put on the clustering algorithm CLARANS. In order to improve the low accuracy, stability and efficiency of CLARANS algorithm when handling of large data sets, this paper proposes an algorithm whose name is GCLARANS based on the spatial grid structure. Back on the Subscriber Management System's technology framework, this paper designed and implemented a digital TV customer segmentation model based on the improved GCLARANS clustering algorithm.This paper does the followings about the above topic:(1) Firstly, this paper describes basic concepts of the data mining technology and Subscriber Management System, analysis the data mining technology how to apply in Subscriber Management System.(2) Secondly, this paper describes the concept of clustering technology, clustering algorithm classification and clustering algorithm evaluation criteria. In order to improve the poor accuracy, stability and efficiency of CLARANS when working with large data collection, this paper determines to improve the CLARANS' performance by optimizing the initial node selection; reduce the computation work in neighbor node select process. The experiments show that the correctness, validity of the GCLARANS algorithm is validated.(3) Finally, this paper proposed a customer similarity model according to customer segmentation theory, and combined with the GCLARANS algorithm, gives a complete implementation of digital TV customer segmentation model. Test and verify the model on the real data collection.The research results of this paper, including data mining algorithms, and data mining systems construction method, not only can be applied to digital TV industry, but also can be applied to other industries. It's had some reference and instruction values for those who are building or plan to building data mining systems.