Research And Application Of Clustering In Telecom Customer Differentiated Reminder Based On Hadoop

With the rapid development of telecommunications industry, user base growsfast. At the same time, the phenomenon of users owing is increasing. How to improvethe efficiency of arrearage reminder as well as customer perceived value is a bigproblem faced with telecom operators. Therefore, data mining techniques is used toanalyze customer’s arrearage reminder information. Delinquent customers areclassified to apply different solutions, in order to achieve differentiated reminder.However, the traditional single clustering has been unable to meet the demand ofmassive data processing, cloud computing could solve this problem and provide anew direction for cluster analysis.In this thesis, customer information of Nanchang telecom is studied for creatingthe telecom customer data object in arrearage reminder pattern provided that currenttelecom arrearage reminder process and data characteristics has been included. Theapplication of cloud computing and clustering in telecom customer differentiatedreminder area is studied as paralleled MapReduce coding method is studied based onHadoop platform. Optimized K-means strategy based on repeated sampling and spacedensity is proposed to weaken the existing of the initial value. Therefore, theclustering result is more stable without the initialization dependence of the clusteringparameter. Optimized K-means algorithm on Hadoop platform (MRDK-meansalgorithm for short) is proposed and realized. Iris and wine dataset derived fromUCI served as the test set to verify the superiority of MRDK-means algorithms whoowns the advantages of accuracy, high efficency and speed-up ratio compared withother clustering algorithms. Data warehouse and management system of telecomcustomer arrearage reminder pattern are designed to provide a series of service suchas customer differentiated reminder with clustering, effect analysis of reminder, costanalysis and telecom maintenance management. Asp.net, Ajax, Web service andmultithreading techniques is embedded to enhance the customer experience. Finally,we summarize the detail and prospect the subsequent task of the study.