USING LEARNING AGENT AND FILTER AGENT TO ENHANCE ACCURACY OF INTERNET SEARCH RESULT

摘要

This paper describes a search engine that using an interesting-learning agent and an information filter agent at the same time to enhance the accuracy of searching. The interesting-learning agent uses improved ID.3 algorithms to learn a user's profile; the filter agent is based on Vector Space Model. Here we use improved ID.3 at beginning, when the number of web pages that an user provides is smaller than 10, we create the decision tree, when the user provides more new web pages, we only add the new features of the new web pages into our decision tree. When the change of decision tree is bigger than a threshold, we recreate the decision tree. By this way, we can enhance the efficiency of interesting-learning agent. In filter agent we use an improved vector space method based on the structure of html. By this way, we can filter the web pages more exactly.