User profile for personalized web search

Different users usually have different special information needs when they use search engines to find web information. The technologies of personalized web search can be used to solve the problem. An effective way to personalized search engines' results is to construct user profile to present an individual user's preference. Utilizing the relative machine learning techniques, three approaches are proposed to build the user profile in this paper. These approaches are called as Rocchio method, k-Nearest Neighbors method and Support Vector Machines method. Experimental results based on a constructed dataset show that k-Nearest Neighbors method is better than others for its efficiency and robustness.