聚类算法在高校网络用户行为分析中的应用

Application of clustering algorithm in university network user behavior analysis

高校网络管理部门在运行管理过程中积累了大量用户上网行为数据,对用户上网行为进行整理分析将能掌握用户上网习惯、规律,科学有效地制定上网管理策略。以一具体高校为例,通过对用户上网数据进行预处理,抽取相应字段构建分析数据集,通过图表形式对上网登录时间进行统计展示。以上网时长为指标值,分别使用K-均值聚类与Kohonen神经网络聚类方法对上网记录进行聚类分析,得到聚类结果。结合用户信息,以用户与上网记录的对应准则作为判断聚类效果的准则,对两种聚类方式得到的结果进行比较,选择合适的结果。结合计算结果对实验单位的上网情况进行分析,对上网管理策略提出建议。

The network management departments in universities have accumulated users′ mass online behavior data in operation management process,which can master users′ online habit and regular pattern by reorganizing and analyzing the users′ online behavior,and formulate the online management strategy scientifically and effectively. A specific college is taken as the example,the users′ online data is preprocessed,and corresponding field is extracted to built the analysis dataset. The online login time is showed in graphic form after statistics. By taking online time as the index value,the clustering analysis for the online record is conducted with K-means clustering and Kohonen neural network clustering methods to obtain the clustering results. In combination with the user information,the results obtained from the two clustering methods are compared by taking corresponding criterion of user and online record as the criterion to judge the clustering effect,and the suitable result is selected. The online condition of the experimental unit is analyzed with the computed results to propose some suggestions for online management strategy.