A GA-based dynamic personalized filtering for Internet search service on multi-search engine

摘要

This paper proposed a methodology based on the genetic algorithm (GA) to provide a novel way for searching via the Internet with a multi-search engine. The proposed system provides a WWW interface for users to enter the keyword. The system connects to the Octopus to get all links and contents. According to the user profile, the system dynamically trains the weights of each keyword by GA, and sorts all the links to adopt 10 best-fit to the user. When the user had browsed the first viewing page and gave the system some feedbacks, the system than utilizes the user's feedbacks to re-train the weights of each remaining keywords. After the training, it uses the new weight combination for the keywords set to determine the order of the remaining links to prepare the next viewing page, and so on. The experiments show that it is suitable for those people that want to collect comprehensive information via the Internet.