Web Usage Mining Applied to Records of Navigation by Internet

Abstract

This paper presents a Knowledge Discovery on Databases (KDD) process applied on the internetsurfing logs at the University of Informatics Sciences. In this context, it describes a Web-UsageMining process using as data sources; the internet surfing logs stored by the proxy server, and alsodescriptive information regarding the users of such surfing service, which was provided by the institution’spersonnel management systems. Statistical, numerical and clustering techniques were combinedseeking to identify user groups with similar internet surfing account usage, in hopes of providingimportant information for decision making processes carried out by the Network Management andSecurity Office or other areas of the institution. This paper describes the methods and techniquesused, and the procedure utilized for performing the descriptive clustering task. This procedure proposesthe use of the CUR matricial decomposition to identify the possible number of groups to identify by thek-medoides clustering algorithm. Lastly, the experiments carried out and the evaluations of the groupsobtained are described and examples of some of the patterns obtained are presented.