Abstract: User profiles, descriptions of user interests, can be used by search engines to provide personalized search results. A query optimization method based on user profiles reasoning is presented. This method creates user profiles by classifying users interests into instances in an ontology knowledge base, and then propagates user preferences to find users latent interests by analyzing the semantic association among the ontology instances. It integrates users current and history preferences to process the search results. A prototype system is implemented and the experimental results show that users latent preferences can be learned accurately and personalized search based on user preference yields significant improvements over the origin results.