Research And Application Of Data Mining In The Model Of Off-grid Users

After the restructuring of China’s telecommunications industry in2008, thetraditional competitive advantages gap between China Mobile、China Unicom andChina Telecom are shrinking, while the competition is increasing. Simultaneouslywith the arrival of the mobile Internet era, the OT（Iover the top）impact on traditionalbusiness communications industry is also growing rapidly, China Mobile’s marketshare is declining. The main factor that causes the mobile market situation graduallybeing eroded is customer loss; there are many reasons for customer churn.However,under normal circumstances, before the loss of customers, their business tends willoccur a change in habits. By monitoring and judging their business habits, thecustomers can be retained by taking care of them in advance.Analysis and predictingof the customer loss is called loss warning.This paper is based on Zhengzhou Mobile’s mobile communication market.The off-grid status and off-grid reasons of users in Zhengzhou can be analyzedthrough data mining analysis of the customer database. The factors which indicate theuser losing can be dug out. Factors include the following aspects: call duration andnumber of calls; call charges; the variation of internet flow; the number of customercomplaints; number of call transfer between different networks. The threshold valueof each factor can be calculated by computational analysis.Based on the decision tree algorithm, using the customer churn behavioranalysis as the premise, its characteristics can be summarized based on the existingoff grid customer data. After quantifying the quit behavior influence factor usingdecision tree algorithm which can automatically determine the customer churn rules,and also determines whether the user has the tendency from the network. The earlywarning model is simple and convenient. It has rapid classification speed and highaccuracy. This model,which is verified to have a good forecast effect in the actualoperation through experimental analysis, play an important role in customer churnwarning.