Information Technology Reference
In-Depth Information
r ela ted work
We briefly discuss Web usage mining techniques and its applications in adaptive Web sites and mobile
Web navigation.
Web usage Mining
Web usage mining refers to the mining of Web server logs to find interesting patterns in Web usage.
Web server logs are preprocessed to find sessions. Conceptually, a session is a single visit to a Web site
by a user. A session is represented by the pages browsed in that visit. From sessions, many Web usage
patterns can be extracted, including associations, frequent paths, and clusters.
Association rules represent correlations among objects, which were first proposed to capture correla-
tions among items in transactional data. If a session is viewed as a transaction, association rule mining
algorithms can be employed to find associative relations among pages browsed (Mobasher, Cooley, &
Srivastava, 1999a; Yang, Zhang, & Li, 2001). Using the same algorithms, we may find frequent paths
traversed by many users (Frias-Martines & Karamcheti, 2002).
If each Web page represents a dimension, a session can be represented as a vector in the page space.
Sessions can be clustered based on their similarity in the page space. In other words, sessions contain-
ing similar pages will be grouped into clusters (Fu, Sandhu, & Shih, 1999; Shahabi, Zarkesh, Adibi,
& Shah, 1997).
Adaptive Web Sites
Perkowitz and Etzioni (1997) proposed adaptive Web site as a solution to the problem of Web navigation.
An adaptive Web site is a Web site that automatically or semi-automatically adapts it structure based
on user browsing. They proposed creating dynamic links using Web usage mining techniques. Koutri,
Daskalaki, & Avouris, (2002) gives an overview of techniques for adaptive Web sites.
Anderson et al. (2001) argued that an adaptive Web site is especially interesting for mobile Web
navigation. Due to limited display size, computing and storage capability, and network bandwidth, Web
sites developed mainly for desktops deliver content poorly to mobile devices. To better serve the needs
of mobile Web users, they proposed building Web site “personalizers” that observe the behavior of
Web visitors and automatically customize and adapt Web sites for each mobile visitor. The MINPATH
algorithm as described in Section 2.3 epitomizes their approach.
The MINPATH algorithm tries to improve the mobile Web user browsing experience by suggest-
ing useful shortcuts. MINPATH finds shortcuts by using a learned model of user behavior to estimate
the savings provided by shortcuts. The shortcuts are dynamically inserted into the page that the user
will browse next. For example, after a user browsed pages A-B-C, MINPATH may provide a shortcut
D->K in the next page D. It uses a prediction model that learns the user browsing behaviors to find the
best shortcuts. If the user follows this shortcut, the user session becomes A-B-C-D-K. Assuming that
without shortcut, the user session would contain pages A-B-C-D-E-F-G-H-I-J-K, the shortcut results
in a saving of six pages or links.
Search WWH ::
Custom Search