CDR Based Temporal-Spatial Analysis of Anomalous Mobile Users

As the competition in mobile markets becomes fiercer, telecom operators have devoted themselves to serving subscribers better. Among large number of subscribers, there are some anomalous users who should be paid more attention to, such as those issuing a lot of unsolicited calls to normal subscribers or making calls to one single phone number continuously. Our study is based on the data of Call Detail Records(CDR) of 10,000 subscribers provided by a Chinese telecom operator, who are randomly sampled from a high-tech industry district for one month. In this paper, we distinguish the abnormal users from the normal ones using agglomerative hierarchical clustering, and divide the anomalous users into telecom marketers or frauds, mobile hot lines and testers. We find that all the anomalous users are distributed in urban areas. In addition, testers and telecom marketers or frauds have relatively weak mobility, and are located in the centralized areas. On the contrary, users being recognized as mobile hot lines show the characteristics of high mobility and relatively decentralized distribution in geographical aspects.