【作者】 刘杰；

【导师】 雷振明；

【作者基本信息】 北京邮电大学 ， 信号与信息处理， 2013， 硕士

【摘要】 伴随着互联网的迅猛发展,网络的规模正在不断扩大,互联网已深入到人们生活的方方面面。在使用各种网络业务的同时,人们在与网络、网络中极为丰富的各种信息以及其他网络用户的交互过程中,表现出各式各样的网络行为。如果说,前些年发展十分迅速的固网宽带给人们打开了通往丰富多彩的互联网世界的大门,使人们接触到网络中多种多样的信息,那么最近几年快速发展的移动网络则使得人们更为方便和简单地接入互联网。有了移动互联网络,人们可以不再像只有固网的时候那样受到一些固有条件的制约,可以更加随心所欲地在任何时间任何地点接入网络,从而更简单便捷地获得所需要的各种信息。本文分析了网络用户行为分析的意义,介绍了用户行为分析的概念、特点,需要分析的内容和使用的步骤,并简单罗列了常用的分析方法。然后阐述了数据挖掘和海量数据处理的基本知识,包括数据挖掘的目的、步骤和方法,以及海量数据处理的难点与处理技巧。接下来简要介绍了RADIUS协议,以及RADIUS报文结构,设计实现了基于RADIUS协议的报文采集及处理系统,重点对该系统的通信方式(包括通信协议、通信过程和通信端口)、帐号信息来源、信息保存和系统的工作流程进行描述,并对系统的一些性能指标进行了估算。在对系统的工作流程介绍中,详细介绍了报文解析、帐号状态维护、帐号超时维护、数据保存和库表维护几个子流程。之后以通过该系统采集的用户上下线记录为源数据,进行网络用户的网络行为特征分析,对ADSL固网用户和CDMA移动网络环境下的会话特征和用户特征进行分析比对,主要包括对上下线时间、会话长度、上线次数、用户在线时长以及流量特征等属性的分析,最后对两种网络环境下的用户利用K-Means算法进行聚类分析,并对分类后的各组表现出的特征进行深入分析和对比。更多还原

【Abstract】 During recent years, with the rapid development of the Internet, the size of the network is expanding. Nowadays, the Internet has penetrated into every aspect of people’s life. It has become an essential part of our life. While using a variety of network services, people in the process of interaction with the network, the extremely rich variety of information the network shows, as well as other network users, there are a variety of network behaviors. If the fix broadband which developed rapidly in the early years lead people open a door to the colorful Internet world, and so that people are exposed to a variety of information in the network. Then the rapid development of mobile network in recent years facilitates people to access the Internet much more. With mobile Internet, people are no longer as fixed-line by the some constraints inherent conditions, so they can access the network at any time and any place to search any information they need.The significance of analysis of network users’ behavior is introduced firstly in this paper. Then the concept and characteristic of analysis of network users’ behavior are discussed. Some common methods of analysis are listed. And then we introduce the basic knowledge of data mining and mass data processing, including purposes, procedures and methods of data mining, as well as the difficulty of mass data processing. In the next chapter, the RADIUS protocol, and RADIUS message structure are briefly introduced. And then we describe the system of message collection and process which is based on RADIUS protocol. The major of the chapter is introduction of work process of the system, and some performance index of the system is estimated. In the introduction of work process of the system, the several child parts including packets analysis, the account status maintenance, account timeout maintenance, data preservation and library table maintenance are showed in detail. In the following part, we use user logs of on and off line which are generated in this system as source data; analyze network user’s network behavior characteristics. We carry on the analysis and comparison of session features and user characteristic for ADSL fixed network users and CDMA mobile network environment, mainly including the attributes such as on and off line time, conversation length, number of users login, duration and flow characteristics. Then we use K-Means algorithm, do clustering analysis for users under two environments, do the in-depth analysis and contrast of feature for the groups after classification. Finally, we prospect aspects which need to improve and expand.更多还原