Page interest estimation model considering user interest drift

A page interest estimation model based on analyzing the users´ browsing behaviors recorded in web access logs is proposed. This model avoids users´ feedback to the browser and doesn´t collect the information which may produce privacy issues. To provide the proper data set for the estimation model, a referer-based data preprocessing method is firstly carried out to improve the reliability of the access data and extract the necessary information for page interest estimation. The computation model is then designed based on the basic idea that the interest of pages to every user is primarily dominated by the reference length. The user interest drift is also considered in the estimation model by using the visiting time of pages to modify the reference length. The developed model is verified by the practical web access logs. The page interest can be efficiently calculated and the data set is well prepared for constructing user profiles.