基于K-均值与AGNES聚类算法的校园网行为分析系统研究  

Research on campus network user behavior analysis system based on K-means and AGNES clustering algorithm

作　　者：茆汉国 

机构地区：[1]南京工程学院信息化建设与管理办公室,江苏南京211167

出　　处：《现代电子技术》2016年第23期116-120,共5页Modern Electronics Technique

基　　金：南京工程学院科研基金项目:基于Web服务的数字校园短信平台研究与应用(QKJB201313)

摘　　要：校园网中的服务器存有海量的用户访问日志文件,记录了校园网用户的访问信息。鉴于此,提出了一种基于聚类算法的校园网用户行为分析技术,设计和实现了数据预处理系统,对日志数据进行一系列的清理、合并,标准化等预处理,使其更好地适应后续的聚类操作。将预处理后的数据作为输入数据,分别实现了三种常用的聚类算法对日志数据进行聚类,然后从聚类准确率和聚类速度两个角度对现有算法进行优化。为了提高聚类准确率,提出了用K-均值算法结合AGNES算法的方法;为了提高聚类速度,在MPICH2平台上设计和实现了并行K-均值算法,实现多机并行分析,最后简单介绍了校园网行为分析系统的应用。The server in campus network has massive user access log files, and records the access information of the cam- pus network users. In view of this issue, a campus network user behavior analysis technology based on clustering algorithm is proposed. The data preprocessing system was designed and implemented. The log data is conducted with a series of cleaning, merging, standardization and preproeessing to suit for the subsequent clustering operation. The preprocessed data is taken as the input data to cluster the log data by means of 3 commonly-used clustering algorithms respectively, and then, the available algo- rithms are optimized in the aspects of clustering accuracy and clustering speed. In order to improve the clustering accuracy, a method of combining AGNES algorithm with K-means algorithm is proposed. In order to improve the clustering speed, the paral- lel K-means algorithm was designed and implemented on MPICH2 platform to realize the multimachine parallel analysis. The ap- plication of the campus network user behavior analysis system is introduced simply.

关 键 词：校园网 行为分析 聚类算法 MPICH2 

分 类 号：TN98-341[电子电信—信息与通信工程] TM417[电气工程—电器]