Article Preview
Top1. Introduction
There is an emergent need for effective informative retrieval in order to satisfy the information need of the user. Research has been done for improving the precision of search results using Personalization of web search in (Kim, Lee, Lee, & Kang, 2010; Leung, Ng, & Lee, 2008; Zhu, Xu, Ren, Tian, & Li, 2007;. Liu, Yu, & Meng, 2004; Peng, & Lin, 2006; Teevan, Morris & Bush., 2009; Zhou, Lawless, & Wade, 2012; Chirita, Firan, & Nejdl, 2007; Psarras and Jose, 2006; Shen,Tan & Zhai, 2005 ;Yin, Shokouhi & Craswell, 2009, Gao et al., 2007, Cui et al., 2003, Billerbeck et al., 2003).
It is found that most of the techniques of personalized web search based on recommendations classify current users to the cluster of user web query sessions who search on the web with the similar intent as that of current user. The selected cluster is used to generate the web page recommendations where the quality of recommendations depends on the accuracy of classification of current user to a cluster of users. (Arzanian, Akhlaghian, & Moradi, 2010; Nasraoui & Petenes, 2003;Chawla & Bedi,2008,Chawla & Bedi, 2007; Bedi & Chawla,2010; Chawla,2012a; Chawla,2012b; Chawla,2013; Chawla, 2014a; Chawla, 2014b). Artificial neural network has been found to perform well for classification problem. The feed-forward neural network architecture is commonly used for supervised learning. Feed-forward networks are often trained using a back propagation-learning scheme. (Vishwakarma, 2012) Artificial Neural Network has been applied in various fields like signal processing, pattern recognition, computer vision, intelligent control, nonlinear optimization, data fusion and data mining, knowledge discovery etc. (Zhang & Wang, 2008).
The main contribution of this paper is which makes it different from other related work based on neural network is the use of hybrid of Back propagation neural network and Genetic Algorithm for effective classification in the domain of Information retrieval since there is no work done which applies hybrid of BP ANN and Genetic algorithm for effective Information retrieval. The advantage of using hybrid of BP ANN and GA is that it performs the effective classification based on neural network learning without any local minimum problem.
Thus the hybrid of Genetic Algorithm(GA) and BP ANN is proposed for classification of user queries to clusters for effective Personalized Web Search(PWS) proposed in (Chawla & Bedi,2007) where web page recommendations to user’s queries are then based on a cluster of users, and not just a single user . The users are grouped into clusters according to certain similarity criteria between their user models and the relevance of a certain document or item to a user is based on the information of other users who belong to the same group in a collaborative manner. The flowchart of the proposed approach is given in Figure 1.
Figure 1. Flowchart depicting the flow of proposed approach of PWS using GA BP-ANN