Research On Telecom Customer Communiation Circle In Customer Identification

Customer identification is an important research topic in the field of telecom customer relationship management.The main research task is to identify different customer categories from a large number of customer.Telecom customer communication network is formed by establishing links among customers,and the links represent customers communication behavior.Customer communication habits can be understood more comprehensive from the perspective of network structure characteristics.This paper studied the customer identification from the perspective of telecom customer communication circle.The main contents of this paper are as follows:(1)The social network analysis method and the community division algorithm were studied.Firstly,the telecom customer call data was used to construct the telecom customer communication network,and the structural characteristics of the network were analyzed from the perspective of social network analysis.Then,fast Newman algorithm was used to divide the customers into different customer groups,which repersent effective customer communication circles.Finally,the customer groups communication characteristics were analyzed.(2)Because of the limitations of the current single attribute recognition algorithm,and the subjectivity of the attribute selection in the existing multi-attribute recognition algorithm,a new recognition algorithm based on multi-objective optimization algorithm was given.This algorithm applied the algorithm NSGA-?,which combined the KPP-Neg and KPP-Pos problems proposed by Borgatti as the objective function,to identify the core nodes in the network.Then,the set of core nodes were re-evaluated to select the higher set of nodes,and compared with the node set obtained by other indicators.(3)The customer characteristics attribute set was obtained by collecting customer data through market.And the set was reduced to obtain the effective feature attribute.Then,C4.5 algorithm was used to constructe the decision tree classification prediction model in the list of potential customers.The characteristic attributes of the potential customers are extracted from the model,and the effective heterogeneous potential customers were selected according to the attributes.The algorithm proposed in this paper was vertified with the call detail data of a telecom company.The experimental results showed that the algorithm is more accurate to identify the effective core customers and potential customers in the telecom customer communications network,and provided reliable information support for the implementation of customer relationship management.