A Modified Algorithm to Handle Dangling Pages using Hypothetical Node in PageRank Computation

Abstract

The information on the web is intensifying day by day due to which there is a bulky amount of information available on the web. With this large amount of information the process of searching has become a complex task. To overcome this problem and to make the information retrieval simple there is a necessity of some efficient search engine. With the knowledge of structure of the web Google PageRank algorithm is developed. A PageRank algorithm places the most important pages at high ranks depending upon the number of inlinks linking to that page. On the web there are web pages which do not have any outlink, these pages are called dangling pages. These dangling pages are removed from computation in PageRank algorithm which is not justifiable. These pages produce the philosophical issues and computational issue. Some approaches were developed to handle these dangling nodes but none of them solved all the issues related to dangling pages. One approach to handle dangling node is developed with the help of hypothetical node. This approach solves philosophical issue related to dangling web pages, but addition of this hypothetical node to web graph increases the computational problem because the number of iteration is increased greatly.

In this thesis a modified algorithm to handle the dangling node with the concept of hypothetical node without increasing computational problem is proposed. From the modified algorithm the dangling pages are not excluded from the computation. The modified algorithm is converged in less number of iteration than existing algorithm. The quality of modified algorithm is proved by running any web graph on both the existing and modified algorithm and then the number of iteration that each algorithm take to converge the algorithm is compared.