【作者】 王东亚；

【导师】 贾卓生；

【作者基本信息】 北京交通大学 ， 计算机应用技术， 2007， 硕士

【摘要】 随着近年来各高校网络建设的不断完善，校园网络的用户数量持续增加，用户的网络行为也变得更加复杂多样。如何对各种复杂的用户网络行为进行分析，从中找出规律来对学校管理者提供决策支持显得尤为迫切。本论文主要从研究校园网络行为的统计特征出发，为学校的网管人员从整体上了解校园网络的行为提供了新的方法。目的就是当校园网络需要改进和优化时，网管人员能够通过对校园网络各方面信息的综合分析，及时了解校园网络整体运行状态，从而可以方便的从宏观上制定解决问题的途径，最终实现网络管理的优化。本论文的主要工作是采用数据仓库、联机分析处理和数据挖掘三种技术对校园计费系统日志信息进行分析处理，充分利用并结合这三种核心技术内在的联系性和互补性，实现了一种新的校园网络行为的决策支持系统框架。通过对每天产生的大量上网信息进行提取、分析、汇总，找出规律性、价值性的知识，寻找校园网络行为的规律，进而建立校园网络管理框架体系结构模型。在此基础上，通过采用适当的数据挖掘算法和分析方法，进行更加深入的数据挖掘分析，实现了部分预测分析模型的构建，并建立了相应的评估机制。更多还原

【Abstract】 With campus networks increasing in size and complexity, the management of these networks is becoming more and more difficult. The current management measure is achieved mainly by studying the network behavior from the area of network traffic, without the decision support data being available to the network managers on a macro scale.The thesis begins with statistical research on the network behavior, giving the network manager a good insight into how the network actually behaves over a period of time. This allows the manager to determine and establish an effective measure of how the campus network operates during that period, allowing the system to be optimized or upgraded where appropriate.The thesis mainly works by taking "on-line analytical processing" warehouse data mining to analysis and handle the campus network system log, making use of the inherent relationship of the three core technology to realize a new decision support system framework of the campus network behavior. This would be achieved by extracting and analyzing day-to-day generated network information to gather regular and valuable system knowledge, in addition to looking at rules for campus network behavior to establish a structured management framework module system. We could carry out further data mining analysis and realize the set-up of a partially predictable analysis module as well as establish corresponding evaluation mechanism by the means of proper data mining calculations and analysis methodology.更多还原