The Researches Of Complex Network And The Applications On The Telecom Churn Analysis

Customer is the strategic resource of telecom carrier. The cost of developing a new customer is triple times higher than retaining one, so more attention has been paid to the customer churn in the meantime of developing new customers. The techniques of data mining and data warehouse are utilized to analyze customer churn from data set about the metrics of churned or churning customers, and the concluded churn patterns are used to match the customers with same characteristics, who will be retained by corresponding marketing strategies .The traditional analytical data scope covered the churned or churning customer's personal characteristics such as age, gender, salary, call times and so on, which just displayed the relationships between the customer and the enterprise and uncovered the relationships among the customers themselves.This paper introduces the telecom social network, which the link between two customers is established by their communications, this network not only displays the communications between customers but also in some ways is a truly social network.We use four types network which are composed by fifiteen fifteen group datasets these are scientific coauthorship network, phone call network, keyword network and random network.We compare the properties among four types of unweighted networks and the weighted networks with the help of high efficency algorithm and high performance computing;Then we choose the degree, clustering coefficient, betweenness and clique as the properties which can distinct stabile customers and easily churned customers.Finally we make a comparison between the traditional customer chum analysis and the new customer churn analyais based on static metrics of complex network. By introducing the static metrics of complex network, we get better effect of the churn prediction.