基于网络社团检测的电信客户细分

Telecom Customer Segmentation Based on Network Community Detection

现有的电信客户细分方法无法发现基于个体间交互关系形成的客户群体。为此,提出基于网络社团检测的电信客户细分模型。考虑网络加权方法对社团检测效果的影响,采用电信企业的通话明细记录构建不同的加权电话呼叫网络,并采用随机漫步模型算法建立基于通话关系的客户细分模型。在细分模型的基础上,结合客户基本信息分析社团的特征,使用网络节点度中心性识别社团中的中心客户。经电信运营商客户通话数据的分析表明,以通话总时长进行加权的网络得到的社团检测效果最佳,可以用于检测客户的关系圈,发现具有领导地位的中心客户,为电信企业客户挽留、精准营销等提供有效的决策支持。

Since the existing Telecom customer segmentation methods are unable to discover the groups formed by interaction relationship between individuals, a network community detection method is proposed for building customer segmentation model on call network. This paper uses telecom call detail records to construct several weighted networks affect community performance, and employs random walk algorithm to build call-relationship-based customer segmentation model. On the basis of segmentation model, it exploits network degree centrality to find central customers. Empirical results on call data of Telecom operation customers demonstrate that community detection of call-duration-weighted network outperforms others, which provide Telecom operators with efficient commercial support.