Personalized Research Based On Agent And Meta-Search Engine

With the development of the network information technology, people have gradually stepped into the era of information overload from lack of information and as a result, traditional search engines cannot meet users'demands for personalized data. Therefore, as an effective solution, personalized recommendation attracts wide social attention. The paper mainly researches the working principle and designs a framework of the personalized recommendation system which is based on the agents and meta search engine. In addition, this paper discusses how to utilize user preference information from conscious and unconscious response to improve the service of the personalized recommendation system.By analyzing the advantages and disadvantages of the currently used explicit and implicit feedback receiving methods, this paper proposes a mixed user preferences obtaining pattern and utilizes improved representation of users'interests based on vector space. This method adds the long-term mark, short-term interest mark and updated time, and interest level parameters based implicit feedback behavior for the method of utilizing TF/IDF to calculate keyword weight, in order to express the user's interest more effectively.The personalized recommendation system adopts the system filtering recommendation algorithm based on item. According to the mixed user preferences obtaining pattern, this paper quantified the no score material preferences visited by users and supplement user evaluation matrix to improve the recommended effect.According to the deep research for user interest model and the adoption of the Agent technology, this paper designed a framework of the personalized recommendation system which is based on the Agent search engine. This framework mainly contained the functions of human-computer interaction, user interest learning, system data management, information search, agent collaboration and so on. By the close and direct collaboration of every module, this system framework effectively combine agent, meta search engine, user interest model and so on together, in order to provide more accurate demand information for users according to the customized personalized model of user behavior.