基于呼叫详情记录的社会角色推测可视分析-《计算机技术与发展》

文章信息/Info

Title:: Visual Analysis of Social Role Projections Based on Call Detail Records

作者:: 蔡梦杰¹ ; 李学俊¹ ; 王桂娟¹ ; 周锐¹ ; 谭博友¹ ; 赵韦鑫¹ ; 吴亚东²; 1. 西南科技大学计算机科学与技术学院,四川绵阳 621010;
2. 四川轻化工大学计算机科学与工程学院,四川自贡 643002

Author(s):: CAI Meng-jie1 ; LI Xue-jun1 ; WANG Gui-juan1 ; ZHOU Rui1 ; TAN Bo-you1 ; ZHAO Wei-xin1 ; WU Ya-dong2; 1. School of Computer Science & Technology,Southwest University of Science & Technology,Mianyang 621010,China;
2. School of Computer Science & Engineering,Sichuan University of Science & Engineering,Zigong 643002,China

Keywords:: call detail records; social role; trajectory embedding; crowdy behavior pattern; user clustering; visualization analysis

摘要:: 城市居民的社会角色感知对城市规划策略制定与城市安全方案设计具有重要的辅助价值,对于后疫情时代疫情的防控具有重要价值。知晓患者用户的角色,可以对用户的接触人群进行更好地分析,做好疫情防控。该文提出了一种结合基站语义和用户时空状态序列的交互式用户社会角色可视分析框架。首先,基于序列数据建模方法,提出了考虑序列顺序的基站嵌入模型 Pos-Cell2Vec 对基站语义信息进行识别;然后,提出一个基于轨迹序列嵌入的用户聚类方法,获得用户聚类结果,进而采用高维可视化方法对基站以及用户的聚类结果进行可视化;最后,基于多视图协同可视分析技术,设计并实现了基于海量通话数据的用户社会角色推测可视分析系统。结合现实数据案例分析结果发现,分析者能够通过该系统结合用户状态序列、用户的通话特征、移动特征以及基站信息,对用户的社会角色进行推测,目前可以通过系统和模型推测出司机、学生以及推销人员等角色。

Abstract:: The perception of the social roles of urban residents is an important aid to the development of urban planning strategies and thedesign of urban safety programmes,as well as to the prevention and control of epidemics in the post-epidemic era. By knowing the rolesof patient users,a better analysis of the user＇s contact group can be achieved for epidemic prevention 　and control. In this paper,wepropose a framework for interactive user social role visibility analysis that combines base station semantics and user spatio-temporal statesequences. Firstly,a base station embedding model Pos-Cell2Vec is proposed based on the sequence data modelling approach to identifythe semantic information of base stations. Then,a user clustering method based on trajectory sequence embedding is proposed to obtainthe user clustering results,and a high - dimensional visualization method is used to visualize the clustering results of base stations andusers. Finally,based on the multi - view collaborative visual analysis technique,a visual analysis system for user social role inferencebased on massive call data is designed and implemented. The results of the analysis combined with real-life case studies show that thesystem allows analysts to infer the social roles of users by combining the user state sequences, their call characteristics, mobilecharacteristics and base station information.