Using SOM to clustering of web sessions extracted by techniques of web usage mining

Abstract

Everyday a huge amount of pages are published on the Web, and, as a consequence, the users' difficulty to locate those that will meet their needs is increasingly bigger. The challenge for web designers and e-commerce companies is to identify groups of users that present similar interests in order to personalize navigation environments to meet those interests. In an attempt to offer that to the countless web users, in the last years, several researches have been done on clustering applied to Web Usage Mining. In this paper, a log file is preprocessed to map the sequence of visits for each user's session. A Session-Path Matrix is used as input to SOM Map and identifying patterns between each session. The results show the similarities between the sessions based on time spent on visited paths and volume transferred. © 2012 Springer-Verlag.