基于改进派系过滤算法的用户通信模型的分群方法

A Method of Grouping User Communication Model Based on clique percolation method

为了对用户的通话记录数据进行挖掘进而完善公司产品,文章提出了一种基于改进派系过滤算法的用户通信模型的分群方法。以用户手机号为节点、用户的通信行为作为节点之间的边建立起传统CPM网络模型。由于传统CPM算法忽略了节点间的关系强度,于是在此基础上引入权重的概念,将用户之间的通话地点、通话时长以及通话频率转化为边的权重,有效的将无权网络转化为有权网络。通过对有权网络应用CPM算法,可以更精确地实现模型的分群。

In order to improve the company＇s products by digging the user＇s call record data, this paper proposes a grouping method of user communication model based on clique percolation method. The traditional CPM network model is established by using the user＇s mobile phone number as the node and the user＇s communication behavior as the edge between the nodes. As the traditional CPM algorithm ignores the strength of the relationship between nodes, so on the basis of the introduction of the concept of weight, the user＇s call location, call duration and call frequency into the edge of the weight, the effective conversion of the network Rights network. By applying the CPM algorithm to the right network, the clustering of the model can be realized more accurately.