Improvement of Page Ranking Algorithm by Negative Score of Spam Pages
Journal Title: Webology - Year 2019, Vol 16, Issue 2
Abstract
There are billions of web pages in the web and the most significant point is how to search these pages regarding their usefulness. Usually a user enters a query into a search engine and looking for the best responses. Search engines work based on ranking algorithms. Ranking algorithms employ web mining methods. Web mining is divided into structure, content and web usage mining. In search engines, related pages with user's query must be listed as a result and to obtain a better results, content mining algorithms may be included. Features such as input and output weight, content weight, spam score, length of URL, the number of related pages and href tags values are considered with a negative score. Proposed algorithms are implemented and compared with well-known algorithm. In this algorithm, Spam score feature is used which is combined with the age of domain and content weight of pages. In PRS (PageRank with Spam score) and PRST (PageRank with Spam score and Time factor) algorithms a better response is achieved than PR (PageRank) algorithm. Obtained results of the other algorithm, WPCRST (Weighted Page Content Rank with Spam score and Time factor) indicates that all measures are improved comparing with PR algorithm and provide better responses. First proposed algorithm combines extracted spam features from “moz.com” with structural features of Web mining. In the second proposed algorithm age of pages is used in addition to the neighborhood matrix and spam score. This feature is used to fix the rich getting richer. Logically this algorithm offers better scores than the previous algorithms. Third proposed algorithm uses weighting methods. In fact, in this algorithm, weight of inbound links and weight of outbound links and content weight of pages are used simultaneously. This algorithm includes web content mining by using content weight of pages. Values of TFIDF and BM25 algorithms are used to obtain the content weight. Obtained results specify that this algorithm has a better ranking than PRST algorithm except for the measure of precision. About 41 percent improvement can be seen in the measure of NDCG and 48 percent improvement for precision while AP has improved 0.8 percent and finally 2.5 percent improvement can be seen in Mean NDCG. In this algorithm, value of TFIDF algorithm is considered as the content weight.
Authors and Affiliations
Ateye Zeraatkar, Hamid Mirvaziri and Mostafa Ghazizadeh Ahsaee
Keywords
Related Articles
- EP ID EP687817
- DOI 10.14704/WEB/V16I2/a187
- Views 304
- Downloads 0
Webgraph connectivity and dynamics: Russian research institutions
This research paper proposes a webgraph dynamics model for research institutes based on a webgraph constructed on a set of instants of time and “return back” through removal of multiple hyperlinks. Analysis of the dynami...
The Role of Responsive Design in Web Development
Responsive design allows software developers to build a Web page that can dynamically adapt to the size of the devices. This development philosophy enables the rendering of Web pages in a fast and optimized way, ensuri...
Application of Web 2.0 Tools in Medical Librarianship to Support Medicine 2.0
It is almost a decade that social networking technology along with its tools such as blogs, wikis, slidesharing/videosharing and photosharing softwares, podcasts, RSS feeds, mashups, folksonomies, and bookmarks has dev...
Wikipedia popularity from a citation analysis point of view
This study aims to provide an overview of the citation rate of Wikipedia since its launch in 2004. It is worth noting that since its inception Wikipedia, the free international multilingual encyclopedia, has been subject...
On Ontology Alignment Experiments
Ontology Alignment is a process for finding related entities of different ontologies. This paper discusses the results of our research in this area. One of them is a formulation for a new structural measure which exten...