科研之友
微信
新浪微博
Facebook
分享链接
摘要
On the internet, commercial messages are automatically posted on the Monitor of a user. If they match a user's taste, they would be for move effective. The authors have developed a dynamic Web advertisement delivery system. It does not need any prior registration of the user's taste. The system we propose here not only grasps the change of a user's taste in the long term, but also predicts the user's current taste. A Markov model and kMER are used to analyze a user's Web site history. A lot of historical data are necessary to make an accurate Markov model. But here we propose a sophisticated method to estimate the users accurate Markov model with a small amount of historical data. We have implemented the proposed method in the Web advertisement system, and have confirmed its effectiveness.
- 出版日期2009-3
