Link-based Web Ranking Algorithms based on Weighted Graph Using Probabilistic Approach

Abstract

The web today plays an important role in the cultural, educational and commercial life of millions of the users. With the huge amount of information available on the web, users typically rely on the web search engines in order to get the most desired and relevant information. A web search engine’s task is to find the most relevant content on the web regarding the query asked by the user. Even if they allow giving the user relevant pages for any search topic, the numbers of results returned are often too big for the user to be carefully explored. Also, the needs of the different users vary as what seems to be important for one user may be completely irrelevant for the other user As most of the users examine the first few pages so the key for user satisfaction is to give the desired results in the first few pages thus there is a necessity to have an efficient ranking algorithm. Therefore the role of ranking algorithms is crucial i.e. select the pages that are most likely be able to satisfy the user’s needs and also bring those results to the top positions. Over the past decade many ranking algorithms have been proposed but there is a little research performed on the measuring the effectiveness of these algorithms. In this research a new method for making the adjacency matrix of the web graph for ranking algorithms is suggested. The rank of ranking algorithms like those of PageRank, HITS,SALSA and Norm (P) algorithms are calculated. From the rank values various performance measures like Mean Reciprocal Rank, Mean Average Precision and Normalized Discounted Cumulative Gain values are calculated and their efficiency is compared with the present scenario of considering the outlink of a particular webpage. In this we also suggest how to calculate relevancy from the document rank.