Research On Telecom Lte Users Churn Algorithm Based On Data Mining

After the reorganization of the China Telecom in 2008, the competition of telecom market has been more and more intense with the development of 3G and 4G. However due to the increasing saturation of the telecom market, telecom operators coming to realize that retaining users in the network will save more costs than developing new users. Therefore, how to reduce the loss of users has become one of the important work.Now data mining has gradually become a widely tool for recognized the loss users of telecom. with the use of data mining, we can fount the law of loss users from a large number of complex historical data of telecom, and then we can retain the users, in order to achieve the purpose of reducing the loss of users. However, many times we focus on the limited of business, charges, complaints, faults and other aspects. This paper proposes LTE internet data which is more close to the users’behavior. Moreover, the data used in this paper is different from the previous behavioral data in units of months, but in units of days of data, and we can easily find out the changes of the user’s internet behavior.In this paper, with the use of LTE users’online changing data, we use the cluster Algorithm like K-Means Algorithm, FCM Algorithm, Naive Bayes Algorithm, BP Neural Network Algorithm, C4.5 Decision Tree Algorithm to analyze the data, in order to discover the laws of losing users.In this paper, the main work includes:1) To extract the LTE online behavior data of telecommunication users, data fields include daily online frequency, daily spending upstream traffic size, daily spending downstream traffic size, daily online time and date.2) To preprocess the extracted data.3) With Matlab tools, we respectively use K-Means algorithm and FCM algorithm to cluster the data, and study the performance of the two algorithms to obtain more accurate characteristics of the loss of users.4) Respectively using naive bayesian algorithm, BP neural network algorithm and C4.5 decision tree algorithm to forecast the LTE online related data of telecommunication users and make comparative analysis, it is concluded that the efficiency and accuracy of three kinds of algorithms. According to the result of evaluation, we find a best solution to the problem and give early warning to the loss of telecom users.In this paper, we combine with the Shanghai Telecom LTE Internet data and data mining, then analyze the reasons for the loss of users and impact indicators. In order to reduce the loss rate of users we will personalized retention programs to the losing users.