17-03-2014, 06:15 PM
INTRODUCTION
Since the search engine Google is successful in
business, whose core is the PageRank algorithm, the
analysis of the web structure is paid great attention to. The
existing methods are divided two categories[1], whose
representatives are the Pagerank[2] algorithm and the
HITS[3] algorithm. They weight the web page by analyzing
the hyperlinks. But, these methods only consider the static
state. That is, when they calculate the web page weight, they
only take the information at that moment, which is
accumulated through the history.
The home pages of 100 web sites are observed
respectively in December 27, 2008, December 29,2008 and
December 31, 2008. The contents include their authorities
and their PageRank weights. Their PageRank weights are
kept. But, some of their authorities are changed. The
changing of authorities shows that the evaluations about the
web pages are changed. But, the PageRank algorithm
doesn’t reflect that. So, the conclusion is shown as the
following:
? The PageRank algorithm mainly calculates the weights
by the static information of the whole web structure
and less considers the local information.
? The results of the PageRank algorithm lag behind the
real life and can’t predict the trend.
The reason why the PageRank algorithm has the above fault
is that it doesn’t consider the time factors. The HITS
algorithm also considers the authorities and the hubs at that
moment only. So, it has the same fault as the PageRank
algorithm.
Some papers have proposed improved PageRank
algorithms[4-6]. But they also take the information at that
moment only. The paper [7] has proposed a new algorithm
to calculate the web page weights, which considers the
interval between the time when the page is created and the
time when the page is queried. But, the result is not correct.
Because, more later the page is created, more higher its
weight is. And, it is not easy to obtain the parameters when
the algorithm is realized.
So, this paper proposes a new method to weight the web
pages, which takes the dynamic information into account.
This method considers the authority changing and its trend,
which can reflect the current and future evaluation about the
pages.
Since the search engine Google is successful in
business, whose core is the PageRank algorithm, the
analysis of the web structure is paid great attention to. The
existing methods are divided two categories[1], whose
representatives are the Pagerank[2] algorithm and the
HITS[3] algorithm. They weight the web page by analyzing
the hyperlinks. But, these methods only consider the static
state. That is, when they calculate the web page weight, they
only take the information at that moment, which is
accumulated through the history.
The home pages of 100 web sites are observed
respectively in December 27, 2008, December 29,2008 and
December 31, 2008. The contents include their authorities
and their PageRank weights. Their PageRank weights are
kept. But, some of their authorities are changed. The
changing of authorities shows that the evaluations about the
web pages are changed. But, the PageRank algorithm
doesn’t reflect that. So, the conclusion is shown as the
following:
? The PageRank algorithm mainly calculates the weights
by the static information of the whole web structure
and less considers the local information.
? The results of the PageRank algorithm lag behind the
real life and can’t predict the trend.
The reason why the PageRank algorithm has the above fault
is that it doesn’t consider the time factors. The HITS
algorithm also considers the authorities and the hubs at that
moment only. So, it has the same fault as the PageRank
algorithm.
Some papers have proposed improved PageRank
algorithms[4-6]. But they also take the information at that
moment only. The paper [7] has proposed a new algorithm
to calculate the web page weights, which considers the
interval between the time when the page is created and the
time when the page is queried. But, the result is not correct.
Because, more later the page is created, more higher its
weight is. And, it is not easy to obtain the parameters when
the algorithm is realized.
So, this paper proposes a new method to weight the web
pages, which takes the dynamic information into account.
This method considers the authority changing and its trend,
which can reflect the current and future evaluation about the
pages.