Web sites are fast becoming the central interaction point in our society. Our aim is to develop an intelligent self-improving web site that can learn from user access patterns and automatically adapt itself to user's interests. We will investigate different ways of exploiting user access patterns to a web site to synthesize and cluster groups of pages commonly accessed together. The most optimized method found will be used in the implementation of the proposed adaptive web site (aWeb).

The proposed adaptive web site is a smart and intelligent site that uses raw usage logs recorded over time, mines and analyzes its contents to find usage patterns, learns user access patterns and adapts itself accordingly. aWeb collects and maintains user log data for every individual user. Data mining techniques will be used to discover the underlying usage patterns, capture relevant user navigational patterns and generate clusters of users. Each cluster represents a group of users with similar web usage behavior and strong cohesive relationship. The accumulated user access patterns, the characteristics of different clusters and suggested changes are used to carry out web improvement and adaptation. The web customization process can be carried out either by human webmaster or via self-improvement schemes.

aWeb uses agent technology to assist the user with navigating the web site (i.e., it provides customized service to an individual user). It can also act as usage-based recommender system. The other main functionality of aWeb is its ability to provide custom real-time web personalization.

This project core technology which will benefit the entire IT infrastructure and in particular R&D workers in industry, government and academia. It improves and enhances the web usage. e-Commerce is one area that will heavily benefit from the outcome of this research work.

aWeb is one of the first attempts to implement quality web; a web that is autonomous with learning and reasoning capabilities. It is also one of the first attempt to build a 'goal-oriented' web sites. The goal prediction and recognition is one of the biggest challenges of this project. aWeb will have build-in capability to measure and optimize the quality of a web site.

Ali A. Ghorbani

Dr. Ghorbani has held a variety of positions in academia for the past 22 years. He is currently associate professor in the Faculty of Computer Science at UNB. Dr. Ghorbani's research originated in software development, where he designed and developed a number of large scale systems. His current research focus is Web intelligence, agent systems and trust & security. With over 10 years experience in high-tech development at major industrial corporations including experience in R&D supervision, he brings strong technological visionary skills and team leadership to FAST ID. Over the last three years, Ali Ghorbani has established a significant presence at UNB. He has published over 40 journals and refereed conference papers and has edited two volumes in the area of software development. Dr. Ghorbani together with two other members of the faculty received CFI fund to establish a research laboratory at the Faculty of Computer Science, UNB. Dr. Ghorbani is the project leader for one of the successful first-round AIF (Atlantic Innovation Funds) projects. His project, “Adaptive Websites”, is valued at $1.05 million over 4.5 years.

Multiagent System Development Kits: An Evaluation
  Elijah Bitting, Jonathan Carter, Ali Ghorbani
  CNSR 2003 Conference > full paper (pdf 216 KB)

Detecting and Preventing Defamation in Mulitagent Systems
  Elijah Bitting, Jonathan Carter, Ali Ghorbani
  CNSR 2003 Conference > full paper (pdf 256 KB)

Toward an Adaptive Web: The State of the Art and Science
  M. Kilfoil, A. Ghorbani, W. Xing, Z. Lei, J. Lu, J. Zhang, X. Xu
  CNSR 2003 Conference > full paper (pdf 284 KB)

• Intelligent and Adaptive Systems Research Group

[Last Revised: 2003 Dec 11]