用SOM网络从移动通信数据中识别朋友关系  

Using SOM to Recognize Friendship from Mobile Phone Data

作　　者：陈蔼祥[1,2] 

机构地区：[1]广东商学院数学与计算科学学院,广州510320 [2]中山大学软件研究所,广州510275

出　　处：《计算机科学》2011年第B10期370-375,共6页Computer Science

基　　金：广东省自然科学基金(10451032001006140);广州市科技和信息化局(10C12140131);广东高校优秀青年创新人才培育项目(LYM10081);广东省大学生创新实验项目(1059210097)资助

摘　　要：从移动电话通信数据中挖掘用户之间的有用信息被认为是富有研究价值的一项工作。利用从94个Nokia6600用户中收集的实际通信记录,设计并实现了一个行之有效的高效分析算法do&del,以构造手机号码对之间的通信模式,证明了do&del算法的可靠性,并给出了算法的理论复杂性和实验观察结果。然后在do&del算法的基础上,给出了基于SOM(Self-Organizing Map)的关系识别系统RRS(Relationship Recognition System),该系统能够在训练好的SOM网络支持下,有效识别双向朋友、单向/普通朋友、一般人等此类用户关系。Data collected from mobile phones have the potential to provide insight into the relational dynamics of individuals.In this paper,using the reality data collected from 94 Nokia 6600 users,an highly effective parsing algorithm named dodel was designed and realized to construct the communication pattern of a pair phone number.The dodel algorithm＇s correctness,theory time complexity and experimental observations were given.With the output of dodel as training data,a Relationship Recognition System building on SOM neural network was designed and realized.The system RSS can recoginze the reciprocal friends,nonreciprocal friends and reciprocal nonfriends.

关 键 词：现实数据挖掘 移动通信 SOM do&del算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]