The Analysis On Prewarning System Of Losing Senior Customers In Shanghai Mobile Based On Data Mining

The competition among telecom market players has become more and more intensive with the development of 3G business.How to maintain the current customers and attract new users at the most is one of the crucial tasks which need to be paid more attention to.The features of losing customers keep changing, due to the consumer psychology as well as behavior are affected by several complicated aspects, such as the campaigns offered by the competitors, the soft landing of charging, and the changes of related policies and rules.For Shanghai Mobile, the senior customer is a key contributor for the revenue. Pareto principle could be expounded that 80% of Shanghai Mobile's profits come from 20% of its senior clients. Losing them will result in the increase of sales cost, the decline of profit and market share etc. Meanwhile, the amount of the loss group is small and currently the average of loss rate is only 1.17%. But the features of them are complicated.Shanghai Mobile estimates the decline of ARPU by its long-term experience. Once the frequency of call forwarding reaches the rate, outbound service will be used. This method seems to be able to keep some clients, while the rate of keeping the senior clients is low. Prewarning system is useful for raising the efficiency of customer segmentation classification.This article employs data mining to classify telecom customer segmentation by analyzing the actual cases. Customer segmentation classification is based on customer behaviors analysis and the decision tree combined with information entropy production. First of all, after the work of data preparation and pretreatment,the algorithm is proceeded based on various kinds of customer behaviors,and the target data has been divided into several kinds of groups among which the behaviors are enormous different from each other while within which the behavior has slight difference.Meanwhile,the experiment results are analyzed to evaluate and estimate the future market trends, so as to provide precise and reliable information to enterprise decision makers as reference for strategy development.Finally,the article presents the design of an analysis system of customer segmentation,which realizes the function of cluster description,cluster analysis and user management.The system is convenient to execute the operation of customer segmentation. By utilizing this algorithm in actual cases, in 2010 the rate of keeping senior customers raises from 92.3% to 96.9%,and the rate of rise is 4.6%. It also means that 92,000 of 2000,000 senior customers have been saved and the rate of accomplishment reaches 42%. In the commercial activities, the success ratio of marketing by using this algorithm is 8 times higher than that without using it. The effectivity of this algorithm could be approved completely.