基于用户通话记录的社区发现算法与社区画像研究

发布时间：2018-04-06 02:13

本文选题：通话记录　切入点：局部社区发现　出处：《浙江大学》2017年硕士论文

【摘要】：我国大规模地普及移动电话和智能终端产生了海量的移动用户历史数据,其中通话记录能够反映移动用户在真实世界中的社会关系,在网络里用户的社交圈被称为社区,通过分析通话记录发现移动用户可能处于不同的社区,比如有亲友社区、工作社区和爱好社区,不同的社区有不同的特征,这些特征被称为社区画像。本文根据通话记录数据重点研究两个问题,一是如何发现通话记录中的社区,二是如何为发现的社区构建画像。针对第一个问题,提出了基于边权重的模块度评测指标,和基于邻集边的局部社区发现算法。针对第二个问题,提出了构建通话记录网络局部社区画像的方法,另外为了从多角度了解用户的通话特征和习惯,本文还提出了从多角度构建局部社区画像的框架。在公开数据集和通话记录网络上的测试结果验证了本文提出的局部社区发现算法的有效性,具体结果如下:(1)对有标签网络,该算法能够发现比较完整的真实社区;(2)对无标签网络,该算法发现的局部社区具有较高的模块度。在通话记录网络上的测试结果表明,本文提出的多角度构建局部社区画像框架能够有效地刻画用户的通话习惯和个人偏好。
[Abstract]:The extensive popularization of mobile phones and intelligent terminals in China has produced massive historical data of mobile users, in which phone records can reflect the social relations of mobile users in the real world, and the social circle of users in the network is called community.By analyzing the phone records, it is found that mobile users may be in different communities, such as communities with relatives and friends, working communities and loving communities, and different communities have different characteristics, which are called community portraits.According to the data of call record, this paper focuses on two problems, one is how to find the community in the call record, the other is how to construct the portrait of the community.In order to solve the first problem, an index of modular degree evaluation based on edge weight and a local community discovery algorithm based on adjacent set edge are proposed.To solve the second problem, a method of constructing local community portrait of call record network is put forward. In addition, in order to understand the characteristics and habits of users from multiple angles, this paper also proposes a framework for constructing local community portrait from multiple angles.The test results on the open data set and call record network verify the effectiveness of the local community discovery algorithm proposed in this paper. The results are as follows: 1) the tagged network.The algorithm can find a relatively complete real community with a high modularity to the untagged network and the local community found by the algorithm has a high degree of modularity.The test results on the call recording network show that the multi-angle local community portrait framework proposed in this paper can effectively describe the user's calling habits and personal preferences.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP301.6

【参考文献】