Clustering Technology Research And Application In Customer Segmentation

In a fast changing market environment,it becomes more and more strongly to the enterprises that customer resources should be the most important element to achieve the victory.The enterprises have transferred from product-centered model to a customer-centered model. With the continuous development and growth of enterprises,more and more customers have been accumulated,but Minority of them can create value for the enterprise,so customer segmentation must be done. Enterprises must develop different marketing strategies for different customers to improve their value contribution for enterprises. Data mining clustering technology to provide technical support for customer segmentation,it can cluster customers which have the same characteristics into one class.This paper studied data mining clustering technology and application of data mining clustering technology in customer segmentation, Major research work are as follows:(1) This paper studied the data mining clustering algorithm,it analysed K-means algorithm based on Division,DBSCAN algorithm based on Density and Meshing Technology. The shortcomings of K-means algorithm was described in detail.The clustering results based on K-means algorithm depends on the selection of the initial cluster center. As the K-means algorithm select the initial cluster centers randomly, so can not guarantee the accuracy of clustering results;K-means algorithm is very sensitive to the "noise" and the isolated point, a small amount of "noise" point would be to cluster large deviation results.(2) This paper proposed a K-means algorithm based on pseudo-parallel DBSCAN algorithm and grid technology(PPDGK). In order to obtain high-quality clustering center points ,the algorithm uses the DBSCAN algorithm and mesh technology to clustered data sets to exclude"noise"and isolated points. In order to reduce Pre-clustering time, the algorithm used a pseudo-parallel technology. Because of the high quality of the initial clustering center, faster convergence of the algorithm.Clustering time of the algorithm was reduced. Simulation results show that PPDGK clustering algorithm in clustering time and accuracy must be better than K-means algorithm。(3) This paper applied PPDGK algorithm to the retail customer segmentation .First of all,this paper introduced the concept of customer segmentation related;Second,this paper described retail customer segmentation system design based on PPDGK algorithm in detail ;Last,This paper implemented this system and analyzed performance of the algorithm in this system and described the customer Segmentation result which based on PPDGK algorithm.In this system,this paper applied separately PPDGK,K-means and DBSCAN algorithm to the same set of sample data . The results showed that PPDGK algorithm can effectively rule out the sample data set of the "noise" points. The segmentation time based on PPDGK algorithm less than other two and The quality of segmentation results better than other two.