Data Mining Research On User Behavior Log In Campus Network

The college is most early applies one of Internet technology construction computer network organizations. With more and more application service increasing in campus network, mass data about network user behavior were created. On the other hand, with the in-depth applications of data mining technology, it is very significant to do knowledge discovery research of campus network using data mining technology. Therefore, it will be an important direction of network research and management to do analysis based on user behavior logs. This paper will explore and analyze the rules in the records of use behavior logs and construct sort models of network users to master and forecast the state of campus network operation so that service quality of users and service consciousness of network administrators can be improved.Massive behaviors are first investigated, in which analysis is carried out from four aspects:time-divided number of online users, user online time, network flow of user and user access destination address. During the analysis of user access destination address, a program of high efficiency is implemented using PERL language to split address segment from mass records of user logs and do sort statistics. The analysis on massive behaviors above can provide good guidance for network administrators to do related work about network bandwidth assignment.Based on the full understanding about massive behaviors of network users, the paper presents useful models for the analysis of campus network user behavior and system analysis about campus network user behavior of Engineering College of Nanjing Agriculture University is achieved. Firstly, the application of data mining technology in user behavior analysis is discussed and the implementation strategy and method for clustering algorithm is presented. Secondly, K-means algorithm is studied in depth and data logs of Charging Gateway server is mined based on K-means algorithm using data mining tools provided by SQL Server2005, in which models are generated using clustering on use models of campus users and analysis and evaluation of these models are also managed. Meanwhile, this paper also tries to forecast and analyze of network flow of different user types based on Time Series Algorithm. In a word, user behavior models above and research work about network flow trend provide reference and guidance for administrator to design network strategy.