User interest correlation in log data
Tao Feng, Pedro Contreras, B. Pauer, Tugba Taskaya, Fionn D. Murtagh · ePrints Soton (University of Southampton) · 2001
When more and more people use web-based information, information of how they use the information is also available in the form of log data. Analysing such data can help information provider to understand their clients’ interests over the information space being served, and adapt it according to users point of view. This paper describes a novel way of applying data mining techniques on Internet logging data in order to find correlated web sections from users’ point of view. We explain how data from the log file can be transformed into a set of transactional click-streams and how data mining techniques can be applied on these transactions. A test bed has been developed for transforming web log data and discovering association rules from it. Real log data from Microsoft web site is used in experiments and evaluation results show that the approach is effective in obtaining useful knowledge of users correlated interests at a particular web site. We also make some effort on mining other service log data obtained from IRAIA project, an information retrieval system serving economical data.