Examining how Users Interact with Hypermedia Using a Neural Network
http://www.lmu.ac.uk/ies/comp/staff/dmullier/icai.pdf

Mullier, D.J.
Proceedings of ICAI
2000

Abstract:

Users of hypermedia systems, including the Web, are known to produce distinctive patterns depending upon what it is that they are trying to achieve with the medium. For example, someone who is seeking specific information produces a different browsing pattern than someone who is browsing generally. However, it is also known that people using hypermedia for similar purposes produce similar, but not identical, patterns. Such information would be useful for a browsing aid, since it would enable the better selection of links, for offering to the user, based upon what the user’s task is. This paper describes the architecture and training of a neural network system designed to recognise hypermedia browsing patterns in a prototype hypermedia environment. A further fuzzy-logic based system, which is used to record trends in browsing patterns, is then discussed in outline. Both systems have performed well in small-scale studies, with both real users and simulated-data. Further, both systems have proved robust to the potentially complex behaviour of users. These issues are discussed further.