Web user behaviors prediction system using trend similarity
Nien-Yi Jan, Nancy P. Lin · 2007
In this paper, a new prediction model without extra information is proposed to construct prediction models for a proxy server prefetcting the suitable pages for adapting the website structure and improving the website performance. The trend similarity is proposed to consider both the page similarity and the position similarity of all pages. A testing websites are established in our experimental environment to evaluate accuracy of our prediction system using page correctness rate and order correctness rate criteria. The experimental results show that the 70% prefetched pages are useful.