Article Preview
Top1. Introduction
Information on the Web is huge and information retrieval of relevant documents for a specific information need of web users is a big challenge for search engines. The search engines retrieve large collection of ranked search results for a specific information need out of which very few are relevant. It is found that relevant documents are present lower in ranking of search results due to imprecise search queries and therefore the precision of search results decreases. Research had been done for web page ranking in order to bring more and more relevant documents up in ranking for the improvement of precision of search results (Peng & Lin, 2006; Selvan et al., 2012; Page et al., 1999; Xing & Ghorbani, 2004; Ding et al., 2001; Jayanthi & Jayakumar, 2011). In (Chawla, 2016) GA was used for optimal web page ranking of High Scent Clicked URLs for personalized web search and experimental results confirmed the improvement of precision of search results. It is realized in this research that rank optimization of clicked URLs using GA can be more effective if trust is used to select the web pages for optimization. The optimal ranking of trusted clicked URLs using GA will retrieve more relevant documents up in ranking and therefore improves the precision of search results.
In this paper a novel approach is proposed using hybrid of GA and Trust for optimal ranking of clicked URLs based on clustered web query sessions. The benefit of using hybridization of Trust and GA together is because of the following reasons .GA is parallel in nature and well suited for solving problems where the solution space is huge and time taken to search exhaustively is very high. The trust value of web pages determines the relevance of web page based on web usage data and is calculated using clustered query sessions. Thus, the use of high trust threshold value will select those web pages which are relevant and satisfy the information need of the users most of the time when recommended in search results.
Thus, an algorithm is proposed for web page recommendation of optimal ranking of trusted URLs using hybrid of GA and Trust for effective personalization of user web search. The entire processing of the proposed algorithm is divided into two phase: Phase I (Offline) and Phase II(Online). In Phase I, the query sessions keyword vector is clustered to group the clicked URLs which satisfy similar information need. The clicked URLs are selected based on trust threshold value in a given cluster and GA is applied on the population of possible ranking of trusted clicked URLs in a given cluster for optimization. Thus, at the end of offline processing, each cluster is associated with optimal ranking of trusted clicked URLs.
During online processing, the input query issued for web search select the most similar cluster for the recommendations of optimal ranking of trusted clicked URLs. The recommendation of optimal ranking of trusted URLs continues till the search is personalized to the information need of the user.
The flowchart steps for the proposed approach of web page recommendations of optimal ranking of trusted URLs using GA is given below in Figure 1.
Experiment was conducted on the data set of user query sessions captured on the web in three selected domains Academics, Entertainment and Sports to evaluate the effectiveness of hybridization of GA and Trust for web page recommender system. The results were compared with PWS(with GA)(Chawla,2016) /Classic IR and the improvement in the average precision of search results confirms that the rank optimization of trusted clicked URLs recommend relevant search results for effective Information retrieval .
Figure 1. Shows the steps for clusterwise GA based optimal ranking of trusted clicked URLs