Combining Customer Attribute and Social Network Mining for Prepaid Mobile Churn Prediction
Combining Customer Attribute and Social Network Mining for Prepaid Mobile Churn Prediction
复制标题
DOI：
--
复制
发表时间：
2013
影响因子：
44.1
通讯作者：
P. Kusuma;Frank W. Takes;P. V. D. Putten
中科院分区：
文献类型：
--
作者：
P. Kusuma;Frank W. Takes;P. V. D. Putten
客户流失，即，在竞争中失去客户是移动的电信中的主要问题。本文研究了将常规表格数据挖掘与社交网络挖掘相结合的附加值，利用客户之间的通信形成的图。我们扩展了经典的表格流失数据集的预测来自社会网络的邻居。我们还扩展了传统的社会网络扩散激活模型的信息，从经典的表格流失模型。实验表明，在第二种方法中，表格和社交网络挖掘的结合提高了结果，但总体而言，传统的表格流失模型得分最高。
Customer churn, i.e., losing a customer to the competition, is a major problem in mobile telecommunications. This paper investigates the added value of combining regular tabular data mining with social network mining, leveraging the graph formed by communications between customers. We extend classical tabular churn datasets with predictors derived from social network neighborhoods. We also extend traditional social network spreading activation models with information from classical tabular churn models. Experiments show that in the second approach the combination of tabular and social network mining improves results, but overall the traditional tabular churn models score best.