An Attribute Relationship Clustering Algorithm for Telecom Customer Group Discovery

This article focuses on the issue of telecommunications customer management, and studies customer group discovery methods based on customer communication relationships and customer attributes. A density clustering algorithm is proposed that considers both communication customer attributes and communication relationships between customers. The customer clustering algorithm is based on an undirected weighted network with integrated similarity. It sorts the nodes in the network according to their strength, starting from the object with the highest node strength. The pruned nodes in the undirected weighted network that have a direct connection to the clustering starting point after removing edges that are less than the similarity threshold are clustered into one group. The starting point of this clustering is the core node, while also obtaining hub nodes and isolated nodes. The empirical analysis of customer data from a telecommunications company in Beijing shows that the customer clustering algorithm proposed in this study can effectively discover customer groups and identify core objects within the customer group.