Increasing search engine's accuracy using links clustering

Abstract

This study is devoted for searching free e-books problem because of its
importance with spreading the e-libraries. When the web client asking for a free
e-books the search engine returns selling e-books sites, this produce a
conclusion of inaccuracy about this search engine. According to this case this
study suggest the philosophy of clustering a sample of web sites on the
number of links which lead to the downloadable e-books and rank this sites
higher than the others with matching the entered keywords. Three clusters
have been noticed (sites with high rank, sites with medium ranking, and low
ranked sites).
The proposed system (Ico.com) has been built form many components; the
first is the web crawler which is a software Program for fetching data of web
pages (like page address, subject, number of links…etc), the second
component is the clustering program written using (VB6.0) language to cluster
the web pages databases according to the number of links to real existing
downloadable books, and the last component is the search engine interface
which is built using HTML and VB script under Active server pages
technology. This work reached to increasing the search engine accuracy using
factors like analyzing the number and the type of links by mining the web site’s
database to improve accuracy for search engines.