| With the fast growth of data available on the World Wide Web, discovery and analysis of useful information from the web has become a practical necessity. Web usage mining, the application of data mining technologies to discover usage patterns from web data, will play an important role in the design and implementation of intelligent web based information systems.;This thesis focuses on the application of data mining in the development of an intelligent agent that makes recommendations to a web site's users based on learning from web usage data and users' behavior. Like a tour guide, the agent assists a user in navigating the web site. Knowledge obtained by the agent may also be used to improve the design of the web site.;A novel algorithm, called LCSA (L&barbelow;eader, C&barbelow;4.5, and web S&barbelow;tructure for an A&barbelow;daptive web site), is introduced to combine web usage mining and machine learning techniques. This algorithm analyzes overall usage, web page content, web site structure, and the current user's actions to provide data for an intelligent agent that supports the user in navigating a web site. We implemented an intelligent agent software system called PagePrompter. Such an agent may improve the performance of a web site, save users time, and increase the profitability of an e-commerce web site. We also implemented a web usage mining system, called WebUsageMiner, which provides the data mining mechanisms to find patterns and relationships from log files and the statistical mechanisms to generate statistics about a web site. Overall, this thesis describes a novel combination of research ideas concerning intelligent agents and data mining. |
