基于局部呼叫数据建模的相对相似用户分布式查询

Relative Similar User Distributed Query Based on Local CDR Modeling

在众多以手机呼叫记录(Call Detail Record,CDR)为数据源的分析研究和挖掘应用中,相似用户查询作为基础研究方法占据着重要地位。传统的查询算法多为集中式处理,然而CDR数据分布产生和存储的本质带来了相似用户分布式查询的问题。本文结合真实数据集,分析用户存储于各基站的局部数据与全局数据的关系,提出并实现基于局部呼叫数据建模的相对相似用户分布式查询方法(RSU-DQ),使用真实数据设计实验验证所提出方法的准确性和高效性。

In many CDR （call detail record） based research and data mining applications, similar user query serves as a funda- mental function. Traditional solutions are more like to be centralized query algorithms, however, the distributed nature of the pro- duction and storage of CDR posed a new problem： similar user distributed query. Based on real world data, we analyzed the rela- tionships between local data and global data, designed and implemented a local CDR-based relative similar user distributed query （RSU-DQ） approach, and then evaluated its accuracy and efficiency.