Better user interest estimation to increase web banner ad click rate

Abstract

Accurate estimation and update of user interest in different categories is necessary so that the web pages or banner ads the user may be interested in are given priority over the other. The web pages viewed in the past or viewing of the popular web pages should affect the user interest less than more recent or less popular page views. In this study, based on tracking of a large number of users, it is found out that the effect of web pages viewed in the past on the recent page views decays exponentially. The effect of giving less weight to popular and old page views on prediction accuracy of both page view categories and banner click categories, is also examined.