Eureka! Institutional Repository

One of the most important challenges that search engines are facing is the evaluation and ranking of the web pages being retrieved for each query. Many algorithms, such as PageRank and Weighted PageRank have been developed for this purpose. The first and most significant one, is being used by Google Search and its concept is that each web page conveys its grade to its links equivalently. On the contrary, the second one computes the percentage of its grade to be conveyed based on the popularity of pages. In this paper, a page ranking method based on the significance of hyperlinks is proposed and implemented experimentally. Each link has a set of attributes, such as its position and size that can differentiate it from the other links of its containing page, thus giving it a specific value. The purpose of the proposed method is to approximate the typical user’s behavior, correlating the probability to follow a link to the link’s relative value. In order to implement this method, there was developed a web bot that collected and stored a sufficient amount of web pages and the attributes of their links. Subsequently, there were applied four algorithms based on the proposed method, along with PageRank και Weighted PageRank. The experiments showcased that this method’s results are clearly closer to PageRank than those of Weighted PageRank.