Clustering Mining In Telecom Customer Classification Of Research And Application

Customer relationship management (CRM) has a vital role to improve the relationshipbetween customers and enterprises. Large customer information to the efficient and accurateclassification is carries on the effective customer management foundation and importanttechnology. Improve telecom enterprise competitiveness will focus on that how to integration,extracts useful data from the no correlation and valueless data in database, and carries on theanalysis research to make differentiation and personalized service.Base on the data analysis, Data mining found the inherent low in l large amount of rawdata. One of the working process is data preparation: select the data from the related databaseand will be integrated into a mining data set which can be used later; The second is found thatrule: use the data which get from collection phase, using a certain way to search and concludethe data set laws. This article mainly aims at an important area in the data mining technology:clustering analysis of research and analysis how to apply it into the reality enterprisedecision-making. The essence of clustering mining is to group those abstract object set, makeit become by many similar objects by the formation of the object class of a process. Therefore,we talk about the most goal of analysis is to collect the similar data classification andpolymerization.This paper refer to a large number of domestic and foreign literature, the data miningtechnology, CRM system, the telecommunications industry customer segmentation model tocarry on the thorough research. Base on the data mining of the customer relationshipmanagement system of customer segmentation.The mainly research work list below:Firstly, set up to data analysis customer classification model. The establishment of the customer classification model will greatly improve the telecommunication enterprises to theuser’s classification, and has some kind of consumption habits or propensity to consume usersummarize up, this thesis mainly is to SMS high frequency using the user class, maketargeted marketing strategies. Model mainly divided into customer data acquisition, datapreprocessing, data mining, and the results visualization output four most. The customerclassification model can be a lot of similar characteristics of the user collection together toform a specific customer category, is the enterprise able to quantify to a user class analysis,and are suitable for telecommunications products and services.Secondly, from two aspects of the original K-means algorithm to optimization analysis.One of the aspects is: the initial clustering center optimization: better classification effect isthrough the initial clustering center choice and get. It makes different clustering of the objectis not similar, and a clustering of the object is similar. At the same time, this paper usemathematical geometry’s law:"triangle two edges must be greater than the length of the thirdside" to reduce the k-means algorithm of total time complexity. The purpose is in order toreduce the number of iteration as far as possible and improving the mining performance.Through the contrast found that the optimized algorithm than traditional K-means algorithmhas better performance.Thirdly, to make sure customer classification model used in the segmentation variables.Consumers itself shows diverse due to the difference between the various factors, Forconsumers, no single strategy can corresponding to the needs of all customers, a singleproduct choice is not a good strategy choice, and customer classification in essence is torapidly improve a large diversified organization management level provides practicalpossibility. Customer classification customer consumption behavior and customer value asthe research variables, customer population characteristics, reference customers psychologicalconsumption factor as references, so as to establish a set of data mining based on customerclassification model.Fourthly, this paper use SPSS Company Clementine7.0as a development tool, base onthe above studies customer classification model, through the telecom customer dataverification, group customer classification successfully. At the same time, this paper selectsout of clustering in the mining results SMS high frequency using group in the analysis ofobject, the consumption habits and propensity to consume are deeply analyzed and formulatethe corresponding marketing strategy. The results show that the work of this paper will be toimprove the telecom customer management, improve customer satisfaction, and improvemarket competitiveness of telecom operators, play a supporting role.