Obtaining User Profiles Via Web Usage Mining.
Marı́a J. Martı́n-Bautista, M.A. Vila, Víctor H. Escobar-Jeria · 2008
ABSTRACT In this paper, we present a model to obtain and analyze user profiles after a process of web usage mining where log files are processed. The web log files register the activity of the user when navigates for a certain web site. These files are mined using a fuzzy clustering process. The obtained clusters represent groups of user sessions which can be related to different demographical classes of people surfing through the web site. These classes are sometimes known and can be identified following the visited web pages. In our case, we obtain and analyzed the user profiles of a University School web server where students and teachers are the usual users of the web site. KEYWORDS Web Mining, Web Usage Mining, Fuzzy Clustering, User Profiles. 1. INTRODUCTION The Web Mining is a researching area that studies the mining of data from the Web. When the type of data to analyze comes from the web log files, the area is called Web Usage Mining. Its objective is to mine web log files to find relations among users regarding their navigational characteristics (Garofalakis et al., 1999). The web mining processes can be optimized by Soft Computing techniques as can be Fuzzy Logic, Genetic Algorithms, Neuronal Network and Rough Sets (Arotaritei and Mitra, 2000). Especially Fuzzy Logic (Zadeh, 1975), helps us to manage the topics related with comprehensibility of patterns, noisy and incomplete data, information of mixed techniques and human interaction. It also gives more flexibility to the systems and produces more interpretable solutions. In Web Usage Mining, the Fuzzy Logic can be present not only in the mining technique, such as Fuzzy Association Rules (Wong et al., 2001) or Fuzzy Clustering (Krishnapuram et al., 2001), but also in the construction of user profiles to manage information such as the user’s age or how patient is the user in her/his navigation, among others (Martin-Bautista et al., 2004). User profiles can be defined as a representation of the knowledge about the user’s interesting information. In (Martin-Bautista et al., 2004), the authors propose two different types of profiles: the simple profiles, which are represented by data extracted from documents supposedly interesting for the user; and the extended profiles containing additional knowledge about the user such as the age, the language level, the country, among others. There are some personalization systems in the literature, although not all of them present real experiments, and the majority of them use association rules. In (Mobasher, 2005), an overview of the process of personalization based in Web Usage Mining is shown. Techniques of data mining such as clustering to discover groups of users are utilized. In this article, we present a model to obtain user profiles after a process of web usage mining using a fuzzy clustering technique. The main contributions of the paper come mainly from three points. First, the proposal of a complete model to obtain user profiles starting from a web usage mining process (section 2). Second, the use of a fuzzy clustering algorithm, that is more suitable for this problem