Studies on mobile agents for query retrieval and web page categorization using neural networks

Abstract

Mobile agent is an emerging technology that is gaining momentum in the eld of distributed computing. There are some advantages in using the mobile agent technology compared with a traditional client-server solution. For example, it can reduce a network traffic, it can support a large scale of computations with many computers in a distributed environment, it allows the use of disconnected computing for processing user queries, and it provides more exibility in the development and maintenance of distributed applications. The goal of this research is based on the application of mobile agent technology in supporting the query retrieval process from the World Wide Web (WWW). Specically, the methods of dispatching the mobile agents to retrieve the query results from the search engines in WWW have been investigated. We have also considered the ranking and classifcation methods applied to the query results that have been retrieved by the mobile agents. The scopes of the research are as follows: First, the effectiveness of mobile agent for query retrieval using the off-line and on-line approaches is investigated. We have found that the query retrieval using the off-line approach by the mobile agent is better compared with the on-line approach. Second, the ranking of query retrieval results that have been retrieved by the mobile agents using the Number of Ordering Score (NROS) is investigated. The Precision of the query results using the NROS is higher than the Recall scores. It indicates that from all of the documents returned from the query, a large proportion of the documents is relevant to the user by using the NROS approach. Third, the performance of mobile agents for query retrieval using an extended hierarchical query retrieval (EHQR) approach compared with the hierarchical query retrieval (HQR) approach is investigated. The result shows that the total routing time taken by the mobile agents to retrieve the query results using the EHQR approach is less compared with the HQR approach. Fourth, the classification of news web pages retrieved by the mobile agents using neural networks based on a background knowledge is evaluated. A new web news categorization approach, namely, a Web Page Classication Method (WPCM) is proposed. The WPCM uses a neural network with inputs obtained by both the principal components and class prole-based features (CPBF). The experimental evaluation demonstrates that the WPCM provides acceptable classification accuracy with the sports news datasets. Finally, we have also overcome the limitation of the principal component analysis-neural networks (PCA-NN method) in supervised data where the characteristic variables that describe smaller classes tend to be lost as a result of the dimensionality reduction by using the WPCM. The classification accuracy on the small classes can be improved although they have been reduced into a small number of principal components.