一体化网络用户行为分析及分类方法研究与设计

发布时间：2018-05-25 10:51

本文选题：一体化网络 + 用户行为分析　；参考：《北京邮电大学》2014年硕士论文

【摘要】：当前开放的互联网架构设计在促进网络蓬勃发展的同时,带来了安全性可信任性差等问题。为此,一体化网络创造性地提出了两层体系结构模型,通过标识分离映射机制,将用户身份信息与位置信息相分离,保障了网络的安全。然而,作为一种新型的网络体系架构,一体化网络在用户行为分析和分类方面还有待完善,尚难以为用户提供个性化和多样化的服务。针对以上问题,本文通过借鉴现有网络中较为成熟的行为分析技术和海量数据处理平台,结合一体化网络自身特点,研究并提出了一体化网络用户行为分析及分类方法。通过分析用户构成和兴趣偏好,掌握用户行为规律,优化资源配置,实现细粒度的管理,来支持丰富的应用服务。本论文的主要工作和成果包括：(1)用户行为分析：包括数据采集、流量分析、内容分析三个部分。具体通过PF_RING高速捕获用户数据包,利用NetFlow读取离线数据包,并通过缓存还原至会话级别,协议分析由IPFIX以插件的形式对数据流处理完成。与此同时,基于MapReduce的分布式平台实现对用户行为内容的分析。(2)用户分类：提出一套用户分类标准,设计用户分类框架结构,结合用户基本属性和行为属性对一体化网络中的用户进行建模。采用基于MapReduce的聚类算法深入挖掘用户上网内容,发现兴趣偏好,实现按行为模式对用户进行分类。最后,通过原型系统平台对上述内容进行了实现和验证,利用Web界面展现了用户行为分析及用户分类结果,包括用户流量统计、用户分布、网站排行、关注热点等。
[Abstract]:At the same time, the open Internet architecture design has brought about the problem of poor security and trustworthiness while promoting the vigorous development of the network. Therefore, the integrated network creatively proposed a two layer architecture model. By identifying the separation mapping mechanism, the user identity information is separated from the location information, and the security of the network is guaranteed. However, as a guarantee of the security of the network, A new network architecture, the integrated network still needs to be improved in the analysis and classification of user behavior. It is still difficult to provide users with personalized and diversified services.
In view of the above problems, this paper studies and proposes an integrated network user behavior analysis and classification method by using the more mature behavior analysis technology and mass data processing platform in the existing network, combining with the characteristics of the integrated network. By analyzing the user composition and interest preference, the user behavior law is grasps, and the resource allocation is optimized. Achieve fine-grained management to support a wealth of application services.
The main work and achievements of this paper include: (1) user behavior analysis, including data acquisition, traffic analysis, content analysis, three parts. The data stream is captured by PF_RING, the offline data packet is read by NetFlow, and the cache is reduced to the session level by caching. The protocol analysis is processed by IPFIX in the form of plug-in. At the same time, the distributed platform based on MapReduce implements the analysis of user behavior content. (2) user classification: put forward a set of user classification standards, design user classification framework, combine users with basic attributes and behavior attributes to model the users in the integrated network. MapReduce based clustering algorithm is used to dig in. In the end, the user behavior analysis and user classification results are displayed by Web interface, including user traffic statistics, user distribution, site ranking, focus of attention and so on.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】