Research And Implementation Of Telecom Customer Churn Warning Model Based On Decision Tree

With the development of mobile internet, telecom industry market booms. Moreand more enterprises join in the telecom industry, and the industry has a largecustomer base. At the same time, the market competition is fierce. Most of theenterprises have customer churn. Analyzing the large number of customer data by themethod of data mining, the enterprise can know the characteristics of customer whichleads the churn. Then, the enterprise can make the right business decisions, to reducethe amount of customer churn. This is the main method of the implementation ofcustomer management.At present, the theory to solve the problem of customer churn is comparativematurity. On the basis of existing research, this paper studies the telecom customerchurn early-warning model by selecting different customer attributes. Clusteringalgorithm is the auxiliary method and the decision tree is the primary method. Thispaper’s study is as follows:Firstly, this paper study the customer’s attributes by using the clusteringalgorithm. Clustering algorithm can deal with large customer data to makeunsupervised classification. This method can give the customer into different groups.Through the analysis for these categories common attributes, then it will get thecommon properties of the lose customer.Secondly, this paper establishes the model of customer churn early-warning. Thedecision tree algorithm is the classification of practical prediction algorithm in datamining. It has particular prediction ability, express to be able to predict the resultsmore intuitive. On the basis of cluster analysis, sorting out users predict erosioncharacteristics of the customers. After selecting the characteristics, this paperimplements the decision tree algorithm by the definition of the data mining process.The first step, clean up the data and obtain the effective data collection; the second step, build a decision tree model; the last step, this paper also evaluate the model, andcalculates the model’s coverage rate and precision rate which shows that the modelhas good properties.Finally, this paper design customer churn early-warning system framework. Byusing the model, new problem will be found. This framework complies with thestandards for data mining and uses the data warehouse technology. This papercompletes the logic design, functional design and module design.This paper chooses the different data characteristics of customer. In addition tothe customer’s information, the Internet traffic is the major characteristics of customer.This paper builds telecom customer churn early-warning decision tree modelwhich has good prediction ability and design customer churn early-warning systemframework which has a good application value.